The $350B Shake-Up: How AI Summaries Rewire Publisher Traffic

How AI Summaries Are Disrupting SEO, Ads, and Publisher Traffic

• Generative AI summaries are upending search: Google, Microsoft (Bing), OpenAI’s ChatGPT, Perplexity and others now deliver instant AI-generated answers in search results and chatbots. These summaries often satisfy users’ queries immediately – leading to fewer clicks on the traditional “blue links” that publishers rely on for traffic pewresearch.org awin.com. The global search ad ecosystem (~$350 billion) is feeling the shockwaves of this shift axios.com.
• Publishers see alarming traffic declines: Multiple studies in 2025 show that when an AI summary appears, users click through far less often. A Pew survey of U.S. Google users found click-through rates on search pages with an AI summary were about half those without (8% vs 15%), and direct clicks on AI-cited sources occurred in just 1% of searches pewresearch.org pewresearch.org. An industry group report (Digital Content Next) linked Google’s new AI “Overviews” to year-over-year referral traffic drops of 10–14% on average, with some content categories and queries seeing CTR plunges over 50% digiday.com digiday.com. In the U.K., publishers documented queries where click-through rate fell from 5% to 0.6% after AI answers were introduced digiday.com.
• Google’s Search Generative Experience (SGE) leads the trend: Google’s SGE (AI Overviews and an “AI Mode” conversational search) now appears on roughly 20% of U.S. desktop queries digiday.com digiday.com. It generates a synthesized answer with citations, often at the top of results. Google asserts that overall search click volume remains “relatively stable” and that AI is prompting more searches and “higher quality clicks” (users clicking through when they truly seek depth) blog.google blog.google. However, publishers and independent data strongly dispute the “no harm done” narrative, pointing to steady erosion of clicks and a surge in zero-click searches (now ~60–65% of all Google searches) pewresearch.org sparktoro.com.
• Bing, ChatGPT, and others adopt AI answers with varying approaches: Microsoft’s Bing Chat (launched Feb 2023 with GPT-4) integrates AI answers into search, always accompanied by footnoted source links. Bing’s market share remains small (~3–4%), but Microsoft claims its AI chat brings in new users and has “a top goal” of driving more traffic to publishers via features like linked citations and even revenue-sharing on chat ads marketingdive.com marketingdive.com. OpenAI’s ChatGPT, by contrast, provides answers based on its trained knowledge (now with optional web browsing), often without requiring any clicks out – effectively bypassing publishers entirely for many informational queries awin.com awin.com. New AI-centric search platforms like Perplexity similarly deliver concise answers with source links, and DuckDuckGo’s DuckAssist offers Wikipedia-based summaries – all signaling an industry-wide move toward answer-first interfaces.
• Publishers fight back with new strategies: To counter AI-driven traffic losses, publishers are experimenting with defensive and adaptive tactics. Many are optimizing content so that if an AI summary uses it, their brand is visible and the snippet teases deeper info – for example, structuring content to encourage clicks beyond the AI blurb. Others are doubling down on E-E-A-T (expertise, experience, authority, trustworthiness) so that their site is favored as an authoritative source in AI results awin.com. Some have erected or tightened paywalls to make their content less freely scrapeable. Industry groups are lobbying regulators to intervene – e.g. UK news publishers submitted evidence to competition authorities that Google’s AI features are causing “irreparable harm” and have demanded an opt-out mechanism that doesn’t torpedo their normal search listing digiday.com digiday.com. In the U.S., publishers are watching the DOJ’s antitrust case, which may force Google to separate its AI crawler from web search indexing, empowering sites to block AI scraping without disappearing from search digiday.com digiday.com. Meanwhile, individual publishers like the New York Times have reportedly taken legal action against AI firms for copyright infringement, arguing that wholesale AI use of their content threatens journalism’s livelihood.
• Ripple effects for advertisers and marketers: The rise of AI answers is reshaping SEO/SEM and affiliate marketing. With fewer users clicking through to sites, organic SEO strategies are in flux – marketers are increasingly focusing on building brand authority and omnipresence (blogs, social, video) so that AI “crawlers” recognize and cite them as trusted sources digiday.com digiday.com. Some brands report reallocating budget from pure search ads to content partnerships and influencer campaigns, hoping to capture users earlier in the research process that AI now mediates. Search ads aren’t disappearing – in fact, Google and Microsoft are adapting by inserting ads into AI results (e.g. sponsored snippets in SGE and Bing Chat), potentially creating new ad inventory marketingdive.com marketingdive.com. But advertisers will need to adjust metrics and tactics as “last-click” attribution becomes murkier. Affiliate sites, especially small “longtail” publishers that rely on search referrals (for product reviews, travel deals, etc.), are under threat: early SGE tests saw some sites’ organic traffic drop ~40% awin.com, which could gut the affiliate revenue of those not commanding unique value or loyal audiences. This is spurring affiliates to improve content quality and niche focus, or diversify traffic sources (e.g. newsletters, community forums), to survive the AI-driven shake-up.

AI Summaries Take Center Stage: Google, Bing, OpenAI & More

In the past two years, AI-generated answers have moved from novelty to a core feature of search engines and digital assistants. Google’s Search Generative Experience (SGE) is perhaps the most consequential implementation due to Google’s dominant market share (~90% globally). Rolled out in beta to U.S. users in mid-2024 and expanded since, SGE produces an “AI Overview” at the top of search results for many queries. This overview is a few sentences to a few paragraphs long (median ~67 words pewresearch.org) and synthesizes information drawn from multiple web sources. Crucially, it includes small citation links to those sources – usually 3 or more are cited for each answer pewresearch.org. Visually, it’s a boxed summary that often contains an AI-generated image or graphic, with clickable source names or an expansion button. Google has also introduced an “AI Mode” (conversational follow-up), allowing users to ask follow-up questions in a chat-like interface blog.google.

Google’s stated approach is to keep users informed of where info is coming from while reducing friction. “Our AI responses feature prominent links, visible citation of sources, and in-line attribution,” Google’s VP of Search Liz Reid wrote, emphasizing that the system is “built to highlight the web” rather than replace it blog.google. In Google’s view, AI summaries help answer novel, complex queries and even drive more searching. Indeed, the company claims users are “searching more than ever” with these new tools, asking longer questions they might not have before blog.google. Google also says SGE surfaces more total links on the results page (between the summary and traditional links below), theoretically giving websites additional opportunities to be clicked blog.google. For simple factoid questions, Google acknowledges an AI box may satisfy the need (“people may be satisfied with the initial response and not click further” – as was already true for things like weather, sports scores, or featured snippets) blog.google. But Google insists that for many queries, users still click through “to dive deeper or make a purchase,” contending that those ensuing clicks are “higher quality” (i.e. users who click are genuinely interested, as opposed to pogo-sticking traffic) blog.google blog.google. In short, Google’s official stance is that overall traffic to the web remains stable, even if user behavior on simple queries is shifting – and the company has a massive incentive to present SGE as a net positive or neutral to the broader ecosystem.

Other search players have implemented their own takes on AI summaries:

• Bing (Microsoft) – Microsoft was first out of the gate integrating GPT-4 into Bing in early 2023. Bing’s search now has a Chat mode that users can engage for a conversational answer, and it also sometimes shows an AI answer on the right sidebar of search results. Microsoft made a point to design Bing Chat with clear citations: footnote numbers in the chat response link to the sources (web pages) from which the information was drawn. Yusuf Mehdi, Microsoft’s consumer CMO, described their philosophy: “first and foremost” drive traffic to publishers even in this new experience marketingdive.com. Early in Bing Chat’s preview, Microsoft added features like a “learn more” section with additional source links, and even an expanded hover-over panel that shows more links from a publisher if you hover on a citation blogs.bing.com blogs.bing.com. The company also began experimenting with ad placements in the chat and said it is exploring sharing that ad revenue with publishers “whose content contributed” to an AI response marketingdive.com marketingdive.com. This was an explicit olive branch to content creators, meant to head off the kind of backlash Google is now facing. Microsoft’s data claimed that the new AI features were adding engagement: by March 2023 Bing had 100 million daily users (with one-third new to Bing via the AI features), resulting in “net new opportunity for publishers” as Mehdi put it blogs.bing.com blogs.bing.com. Still, given Bing’s single-digit market share, the overall impact (positive or negative) on publisher traffic from Bing’s AI is limited in scale compared to Google. But Bing’s approach – active collaboration and revenue-sharing with news partners (7,500+ publishers in its Microsoft Start program) – contrasts with Google’s more unilateral rollout. It suggests one possible model where AI search and content publishers coexist with a more direct value exchange.
• OpenAI’s ChatGPT – ChatGPT is not a traditional search engine, but its astonishing popularity (reaching over 100 million users and becoming a household name in 2023) means it often serves as a proxy for search. Users ask ChatGPT questions they might otherwise have Googled. By design, ChatGPT’s answers are drawn from its training data (a vast corpus of web text up to its knowledge cutoff, plus optional updates) and it does not automatically cite sources. In effect, it can give you an answer sourced from, say, a Wikipedia article or a news site without any attribution or click for the publisher. In 2023, ChatGPT plugins and the Bing-powered Browsing mode became available to allow the bot to fetch real-time information. Even then, unless specifically prompted for sources, ChatGPT typically presents a synthesized answer as if it were the author. This “answer without a click” paradigm has huge implications: if millions of users get cooking recipes, coding help, or travel tips via ChatGPT’s one-stop interface, those are searches and clicks not happening on Google or publisher websites. For news, ChatGPT’s impact was initially blunted by its lack of up-to-date knowledge (and factual accuracy issues), but as of 2025, OpenAI’s tools are rapidly improving at providing current information. Some publishers have grown concerned enough that by mid-2023, major news organizations like the New York Times explicitly disallowed OpenAI’s web crawler (GPTBot) from indexing their content for training, and later pursued legal action against OpenAI for unauthorized use of their articles in AI outputs awin.com culawreview.org. In response, OpenAI struck a deal to license news content from the Associated Press – a sign that even purely AI interfaces may eventually involve paid content partnerships reuters.com. Still, the default ChatGPT model represents the extreme end of the spectrum: an AI answer engine that largely bypasses the web, drawing knowledge from content creators without necessarily returning them traffic or revenue. It’s a scenario that alarms publishers, who see it as the clearest threat to the traditional “search -> click -> ad impression” cycle.
• Perplexity.ai and other AI search startups – A number of smaller platforms are also experimenting with “answer engine” formats. Perplexity is one notable example often praised for its transparent sourcing. It answers user queries by generating a concise summary and listing a set of relevant sources (with direct links) below the answer. In essence, Perplexity acts like an AI meta-search, querying sources (including Bing’s index, Wikipedia, academic papers, etc.) and giving the user a quick digest with citations. This design encourages the user to click through for details – it’s not trying to keep you on the page longer, but rather to save you time finding which sources to read. However, Perplexity’s user base is tiny compared to the big players, so its impact on publishers is correspondingly small (in fact, some publishers might see extra referral traffic from curious Perplexity users). Other search alternatives include You.com (with an AI chat mode and summarized results), Neeva (an AI-driven search engine that was active until mid-2023 and emphasized ad-free, subscription search with AI features before shutting down), and DuckDuckGo’s DuckAssist, which offers to answer simple questions by quoting Wikipedia text. Notably, DuckDuckGo limited its AI summarizer to Wikipedia and a few other sources as a cautious approach – thereby avoiding direct conflict with news publishers for now. This highlights a key divergence: some upstarts see selective or domain-limited AI answers (e.g. only summarizing from public domain or licensed data) as a safer path that won’t cannibalize the broader web. By contrast, Google and Bing are going full steam ahead summarizing “everything” on the open web (unless a publisher blocks them), since they have indexes and AI models sophisticated enough to attempt it.
• Emerging Big Tech moves (Apple, Meta, Amazon) – While not as far along, other giants are angling to incorporate AI summaries too. Apple has reportedly been working on an AI-enhanced search for Siri, even testing Google’s Gemini AI model for “World Knowledge” Q&A in iOS facebook.com. If Apple were to launch its own search interface with AI answers (and rumors suggest an Apple search engine project in the works), that could further disrupt the search landscape – especially given Apple’s control over the default search on iPhones and its recent forays into ads. Meta (Facebook) is also investing in generative AI; while its focus is more on chatbots and feeds, the lines between social platforms and search are blurring (for instance, users might ask an AI inside Instagram for recommendations instead of Googling). Amazon, with its shopping search dominance, has introduced AI features in Alexa and is likely to integrate generative summaries for product Q&A – potentially cutting out third-party review sites from the customer journey. In sum, AI summaries are becoming a ubiquitous feature across tech platforms, not just traditional search engines, meaning the pressure on publisher traffic is coming from multiple directions.

Publishers Lose Ground: The Traffic Freefall and “Zero-Click” Phenomenon

For online publishers – from news outlets to niche bloggers – these AI-driven answers are causing a major referral traffic crunch. The early data from 2024 and 2025 is sobering. When an AI summary appears, user behavior shifts dramatically toward not clicking out to the source websites.

Figure: Pew Research Center found that Google users are far less likely to click a result when an AI summary is present (left), versus when there’s no AI summary (right). In March 2025, only 9% of Google searches with an AI summary led to any click on a result (8% on a traditional result link, 1% on a link inside the AI box), compared to 15% of searches when no AI summary was shown. Meanwhile, 26% of AI-summary searches ended the session immediately (user satisfied or giving up), versus 16% for standard searches pewresearch.org pewresearch.org. This illustrates the “zero-click” impact of AI answers.

Multiple reports confirm the pattern behind those percentages. Pew Research Center’s study (July 2025) tracked 900 U.S. users over a month and found that about 18% of their Google searches generated an AI overview pewresearch.org. On those AI-augmented searches, users clicked significantly less. Only 8% of visits saw a click on a regular search result, and a negligible 1% included a click on one of the AI summary’s cited links pewresearch.org pewresearch.org. In most cases, users either continued refining their search on Google or just left without clicking anything. Notably, users were actually more likely to end their browsing session entirely after seeing an AI answer (26% of searches) than they were on a traditional search page (16%) pewresearch.org. This suggests that the AI snippet often fully answers the query – the user feels no need to go further. Pew’s data also showed an uptick in truly zero-click outcomes: a larger share of users simply closed Google after getting the answer, rather than even continuing to another website or a new search.

This trend compounds a longstanding issue: Zero-click searches have been rising for years (even before generative AI, Google’s featured snippets, knowledge panels, and direct answers meant fewer clicks to external sites). But AI summaries turbocharge that effect. Rand Fishkin’s SparkToro analysis in mid-2024, using clickstream data, estimated that roughly 64% of Google searches in the US ended without a click to any external website sparktoro.com. In the EU, post-regulation, it was a bit better (around 62% zero-click). Those figures include all searches (many of which might be simple one-word queries or navigation), but the trajectory is clear: more queries resolved on Google’s own properties or results page, fewer referrals out. Fishkin noted that by 2024 Google was sending only about 360 out of every 1,000 searches to any open-web site in the U.S. sparktoro.com. That leaves the majority either ending with no external click or going to Google’s own services (YouTube, Maps, etc.). Generative AI answers threaten to tilt that balance even further toward zero-click.

Publisher traffic metrics in 2025 underscore real declines coinciding with AI rollout. Digital Content Next (DCN), a trade association of major publishers (NYTimes, Vox, Condé Nast, etc.), collected data from 19 member companies in May–June 2025 to gauge Google referral trends. The results: the majority of sites saw Google search traffic drop 1% to 25% year-over-year in that period digiday.com. When averaged, the YoY decline in Google search referral was about –10% overall (–7% for news publishers, –14% for non-news publishers) digiday.com. Remember, this is in a world where one might expect search traffic to grow each year; instead, it’s shrinking for many. DCN’s CEO Jason Kint directly blamed Google’s AI Overviews for these drops, citing publisher feedback that the timing lines up with SGE’s expansion digiday.com. “This data offers a ‘ground truth’ of what’s actually happening, cutting through Google’s vague claims about ‘quality clicks’,” Kint said, referring to Google’s public insistence that traffic wasn’t hurting digiday.com. The DCN data suggests that even if overall search volumes are up, the portion of traffic making it out to publishers is down.

The effect seems to vary by content category and query type. Some anecdotal and early evidence: tabloidy or quick-answer content appears hardest hit. In Europe, executives from MailOnline (Daily Mail) reported 50%+ drops in Google referral traffic for certain periods after AI answers rolled out digiday.com. In DCN’s study, non-news publishers (like reference sites, how-to guides, entertainment content) actually saw larger median drops (–14%) than news publishers (–7%) digiday.com. This might be because a lot of non-news queries (e.g. “how to get rid of pantry moths” or “best budget smartphone”) can be directly answered by an AI summary pulling from various sources. Indeed, the UK’s Professional Publishers Association (PPA) provided a striking example: for the query “how to get rid of [an insect]”, a lifestyle publisher member still ranked on page 1 of Google and saw steady impressions, but its click-through rate plummeted from 5.1% to 0.6% over the past year digiday.com. Users got their pest control answer from the snippet and stopped clicking the actual article. Another example: an automotive site ranking #1 for a topic saw a 25% drop in traffic even as its search visibility (ranking) improved – its CTR fell from 2.75% to 1.71% digiday.com. These cases exemplify the decoupling of ranking and clicks in the AI era: you can be number one in organic results, but if an AI box above you satisfies the user, many won’t bother clicking your link.

It’s not just Google web search – the zero-click dynamic extends to other surfaces. Publishers note that Google has also added AI-generated summaries into Google Discover (the mobile feed) and into Google News/Google’s news feed, where articles might now appear as just a citation within an AI blurb digiday.com. Similarly, Bing’s chat can often answer a question that would have otherwise driven a click to a third-party site (though Bing at least shows the link if users want to learn more). Even voice assistants (Alexa, Google Assistant) – essentially AI Q&A – have trained users to expect direct answers without any “see more” link. All this habituates users to get information without clicking through to source sites.

An interesting nuance: AI summaries often draw from certain high-authority sites more than others. Pew found that the most-cited sources in Google’s AI overviews were Wikipedia, YouTube, and Reddit pewresearch.org. Together those three accounted for 15% of AI summary source links (similarly 17% of links in traditional results) pewresearch.org. Government (.gov) sites also appeared more in AI answers (6% of AI sources vs 2% of normal results) pewresearch.org. Meanwhile, news sites constituted only about 5% of links in AI overviews (same as their share in normal results) pewresearch.org. This suggests AI tends to rely on reference-style and community-driven content (which is often evergreen or factual) for answers, rather than news articles (which might go out of date or present narrative content). So while news publishers are absolutely seeing impact (especially on evergreen articles or explainers they publish), informational sites like wiki-how, forums, and reference pages might be even more directly cannibalized. Reddit, for example, has a trove of user-generated Q&A – AI models love to train on that content for conversational answers. Reddit’s leadership famously complained in 2023 about AI companies using its data for free, leading to API changes. This underscores that the entire open web knowledge base is feeding these answers, and not all content creators are willing participants.

From the user perspective, many are happy with AI summaries – they save time. But there is also a risk: if AI answers discourage clicks, users might miss context or multi-perspective views they would get from reading full articles. Some observers worry about the accuracy of AI summaries (early on, both Bing and Google’s AI had some high-profile mistakes, or “hallucinations”). If users trust the summary blindly and don’t click sources, misinfo could spread more easily. However, by Q3 2025 these models have improved, and user trust seems to be growing. Over one-third of U.S. adults now use generative AI tools regularly, and younger demographics have over 50% adoption digiday.com digiday.com. This generational shift suggests upcoming cohorts may be even less inclined to browse multiple sources – they’ll expect the AI to aggregate for them.

In summary, publishers are experiencing a real downturn in the traffic that traditionally came via search engines, especially from Google. AI summaries accelerate the “Google Zero” scenario (a term some publishers use for a future where Google Search sends almost no traffic out). The numbers range from moderate single-digit dips to jaw-dropping collapses for certain query types. And while Google insists things are fine, publishers see the writing on the wall: fewer eyeballs on their pages, fewer ad impressions served, and a need to adapt fast or perish.

Publishers Strike Back: Adaptation, Paywalls, and Legal Pushback

Faced with the loss of traffic – and by extension, revenue – publishers aren’t sitting idle. A mix of strategies is emerging across the industry to defend against or adapt to the AI summary era.

1. Lobbying and Legal Action: Major publishers and media coalitions are pressing regulators and exploring lawsuits to force a more favorable arrangement. In the UK, the Professional Publishers Association (PPA) has formally urged the Competition and Markets Authority to scrutinize Google’s AI search features digiday.com. Their argument is that Google’s 93% search market share, combined with AI overviews, constitutes an anti-competitive practice that diverts traffic from publishers to Google’s own interface digiday.com. A key grievance is the lack of meaningful opt-out: currently, if a site wants to avoid being scraped for AI answers, the only sure way is to block Google entirely (since Google’s crawler dual-purposes content for both traditional indexing and AI) digiday.com. Publishers call this a “Hobson’s choice” – accept our content being possibly used in AI answers (with minimal traffic return), or disappear from Google search results altogether, which is not a viable option for most. Regulators are considering whether this dynamic warrants intervention. In the U.S., the Department of Justice’s ongoing antitrust case against Google has opened the door to potential remedies. Notably, one proposal on the table is to require Google to separate its AI content crawler from its general search index digiday.com. If enforced, this could allow publishers to say “yes to being indexed on Google, no to being summarized by AI.” Jason Kint of DCN said such a remedy would be “big news for publishers” and is urgently needed because “traffic is down double digits… that harm is happening in real time.” digiday.com. He noted a judge could even impose an injunction to force Google to offer that choice before the lengthy case concludes digiday.com. In Europe, publishers are likewise pushing for swift action to give them more control digiday.com.

Beyond regulatory channels, there’s the prospect of direct legal challenges on intellectual property grounds. Late 2023 saw reports that The New York Times and other publishers were considering lawsuits against OpenAI (and possibly Microsoft/Bing) for copyright infringement – essentially arguing that using their articles to train AI or to generate answers exceeds fair use and undermines their business. By mid-2025, it’s reported that The New York Times did file suit against OpenAI culawreview.org. Similarly, a host of authors and content creators have filed class-action lawsuits against AI companies for unauthorized use of their work. While the legal theory is untested, the pressure might result in settlements or licensing agreements. Indeed, OpenAI quickly struck licensing deals with some content providers (e.g., Associated Press, the Financial Times, and other outlets) to legally use their archives reuters.com deeplearning.ai. These deals are relatively small-scale so far (AP’s was described as a limited archive access), but they may set precedents for compensating content originators.

Meanwhile, news publishers in Canada and Australia have leveraged legislation intended to make tech giants pay for news usage (though those were aimed at link aggregation, not AI). It’s possible similar frameworks could be amended or extended to generative AI’s use of news content. In short, the message from publishers: if you’re going to summarize our work, we want a cut. They see AI answers as creating a “free rider” problem where their content fuels another’s platform without compensation.

2. Blocking and Control of Content: Some publishers have taken a hard line by technically blocking AI crawlers. In August 2023, OpenAI introduced a “GPTBot” user agent and offered websites the option to disallow it in robots.txt. A number of prominent sites – including news organizations – promptly did so awin.com. However, this only stops future model training; it doesn’t prevent current models (already trained on past data) from spitting out answers based on that content, nor does it affect Google/Bing AI which use their own crawlers. Google did offer a partial measure: a meta tag to opt out of AI summaries while still allowing indexing. Using nosnippet or limiting snippet length can, in theory, prevent Google from showing an AI excerpt from your page digiday.com. But publishers point out this is a blunt tool – it also removes your normal search snippet, hurting your SEO. Google’s help documentation says you can allow indexing but set max-snippet length to zero or a low number to curtail how much of your text can appear in features like SGE digiday.com. A few publishers have tried aggressively short snippet settings as a test, essentially saying “Google, you can index us for ranking signals, but you’re not allowed to display more than [for example] 50 characters of our content.” This might reduce the usefulness of any AI summary (since the model won’t have enough to work with for that site), but it’s an imperfect solution and could lower click-through if users can’t preview your content.

Some publishers have started adding embedded branding or references within their content in hopes that if an AI summarizes it, the brand name survives in the summary (e.g., writing “According to PublisherName, …” in the text). However, AI models might omit such attribution unless explicitly instructed to preserve it. It’s a bit of a cat-and-mouse game.

On the flip side, a few publishers are experimenting with feeding content to AI systems in a controlled way. For instance, structured data or schema markup can highlight certain facts (think of recipe sites providing structured ingredient lists). The idea is to influence what the AI shows so that maybe it includes a call to action or more context that entices a user to click. There’s even talk of publishers creating “summary friendly” versions of articles – concise answers that Google could use, provided they link back for detail. In essence, if you can’t beat ’em, join ’em: give the AI a good answer to display along with a reason to visit your site. It’s a strategy similar to how some sites optimized for featured snippets in the past.

3. Embracing Paywalls and Exclusive Content: One clear adaptation is to differentiate what’s free on the web (and hence scrapeable) versus what’s exclusive. If the content that an AI summary uses is freely accessible text, then putting more of your premium content behind paywalls or login barriers might protect it (AI can’t easily access paywalled text, at least not without subscription agreements). We’re seeing some publishers tighten paywalls in response to AI. For example, some news sites that used to allow a few free articles are dropping the meter count or requiring login for all content. The calculus is: if we’re not getting the casual Google traffic anyway (because of AI answers), we might as well force the more committed readers into subscriptions. This could accelerate the media industry’s shift to subscription models over ad-driven models, particularly for quality journalism.

Additionally, publishers may create more interactive or non-text content that AI can’t simply scrape and summarize. This includes podcasts, videos, infographics, or even interactive databases. An AI might tell you the highlights of a sports game, but it can’t replace watching the highlights video, for instance. Similarly, some sites have floated the idea of deliberately fragmenting content – not putting the full answer in one neat paragraph. If the writing is more narrative or requires reading multiple sections to get the full context, an AI might have a harder time giving a one-shot answer (although advanced models are quite good at scanning long text). There’s also talk of using watermarking or hidden signals in content that could trigger AI not to use it without credit – though no standard exists for that yet (there’s a proposal for an HTML tag like <ai-nosummary> or similar, but it’s not implemented broadly).

4. SEO Adjustments and “GEO” (Generative Experience Optimization): Despite the upheaval, a core piece of publisher strategy remains search optimization – but now with a twist. Some have started referring to “GEO” – Generative Experience Optimization, meaning optimizing content so it’s featured (and credited) in AI summaries. In practical terms, this involves many classic SEO best practices: ensuring your content is high quality, authoritative, and answers questions directly (so that the AI considers it a relevant source). Jess Sholtz, a publishing consultant and former CMO who led AI strategy at a media group, advises that the fundamentals of SEO shouldn’t change. “If you’re not in the search index, you can’t be in the AI summary citation,” she notes digiday.com. This is a reminder that AI like Google’s is retrieval-based – it largely pulls from the existing search index (using techniques like RAG, retrieval-augmented generation) digiday.com. So publishers still need to do all the things that get them indexed and ranked (site speed, keywords, mobile friendliness, etc.). Some publishers have resisted any drastic SEO overhauls, calling the panic over AI a “hype” they’re not falling for digiday.com. They continue focusing on content quality, knowing that if they rank well and the AI summary cites sources, they have a chance to be one of them.

That said, others are being proactive in tweaking SEO: experimenting with snippet lengths, crafting FAQ sections or “People Also Ask” style content in their pages, which might feed the AI answers. For example, a travel site might add a Q&A section at the bottom of an article (“Q: What’s the best time to visit Paris? A: …”) hoping that if someone asks the AI that, it will pull their snippet and cite them. Some are also paying attention to schema (like marking up reviews, recipes, etc.) which Google’s AI might use to structure answers (Google has hinted that certain schemas might be utilized by SGE for more detailed responses, e.g. pros/cons from reviews). The overall goal of GEO is to remain visible on the AI-dominated results page – either as a cited source or by having content that’s essential enough users will click through for full details.

One emergent insight is the renewed importance of branding. In the old SEO world, being rank #1 was often enough to get clicks, even if the user didn’t recognize your site – there was trust that “if Google puts it first, it must be good.” But in an AI answer, your link might just be a tiny citation with your domain name. The user sees the answer without necessarily knowing who authored it. As Jess Sholtz observed, “On these new AI surfaces, being the top citation doesn’t have that same ‘oh it’s ranking so I trust it — it’s fine’ factor. Branding is that first pillar… someone isn’t going to blindly just click that top citation.” digiday.com In other words, if the user doesn’t recognize or trust the source, they might just take the answer at face value and not bother clicking at all. This is driving publishers to invest more in brand awareness and loyalty. If a user sees an AI summary with a citation from, say, Healthline.com, and they’ve heard of Healthline and trust it, they may be more inclined to click “Read more” to get details or verify. If they see a citation from a random site they’ve never heard of, they might not click – or worse, question the answer’s credibility (though often users just trust the AI or the fact Google presented it). Therefore, publishers are pushing their brands out on other channels – social media, newsletters, podcasts – to build that recognition. Some, like Sholtz’s team, reallocated resources to branding and tech for their smaller titles, recognizing that in an AI-dominated discovery environment, a strong brand could be a deciding factor in capturing the user’s click digiday.com digiday.com.

5. New Content and Engagement Tactics: Losing search traffic forces publishers to diversify how they reach audiences. Many are turning back to what we might call community and direct engagement. For instance, some news and magazine publishers have begun to cultivate networks of content creators or influencers to extend their reach on platforms like TikTok, YouTube, or Instagram – places where AI summaries aren’t competing (yet). Others are rediscovering forums and aggregators: the Digiday podcast noted publishers “looking to places like Reddit for both community engagement and referral traffic”, identifying relevant subreddits to share content or participate in conversations digiday.com. This is somewhat ironic, as Reddit itself is a source for AI answers; but if publishers can get users from Reddit, that’s one step removed from search dependence.

Email newsletters have also seen a renaissance – by cultivating a direct subscriber list, publishers ensure they can push content to users without going through search or social algorithms. We’re also seeing publishers experiment with on-site AI tools to keep users around. For example, some news sites have deployed AI chatbots fine-tuned on their own content archives, letting readers ask questions like “What’s the latest on topic X?” and getting answers drawn from the site’s articles. The hope is to offer an AI-driven experience within the publisher’s domain, rather than ceding that role to Google or ChatGPT. The Wall Street Journal, for one, has reportedly strategized around a “Google zero” scenario by emphasizing its direct relationship with paying readers and offering new tools in its app to keep them within the WSJ ecosystem digiday.com.

Additionally, a few forward-looking publishers are trying to turn AI from foe to friend by negotiating partnerships. We might soon see certain publishers license structured content feeds to AI platforms in return for fees or prominence. For example, a cooking site could partner with an AI assistant such that the assistant preferentially uses their recipes and in return always suggests “From ___, click for full recipe.” This is speculative, but if traffic drops enough, publishers will seek compensation either through deals or via courts.

In summary, the publisher playbook is evolving rapidly: fight where you can (legally and via standards), adapt your SEO and content strategy, and invest in channels that AI can’t take from you. Not all publishers have the resources to do all this. Smaller sites are especially vulnerable; some might simply shut down if their Google traffic dries up. Others might consolidate – we could see a “culling of publishers,” as some experts warn, where only the strongest brands or those with unique value survive digiday.com. This raises alarms about less diversity in online information, but it’s a real possibility if the current trends continue.

Fallout for Advertisers, Search Marketers, and the Affiliate Ecosystem

The shake-up in how users get information doesn’t only affect content publishers; it’s also rewiring the strategies of advertisers, SEO professionals, and affiliate marketers. After all, if user behavior shifts on search platforms, the whole marketing calculus changes with it.

Search Advertisers & SEM: Traditionally, search engine marketing (SEM) involves bidding on keywords so your ad appears alongside organic results. If AI summaries reduce organic clicks, one might assume users would also interact less with ads (since ads are often the top “link” clicks). However, the picture is complex. Google and Bing are not going to sacrifice their cash cow – instead, they are adapting ad formats to the new AI interfaces. Google has already begun testing ads that integrate into the SGE results. For example, when a user gets a generated overview for a shopping-related query, Google might show sponsored product listings or text ads within or just below the AI answer box (clearly labeled as ads). Early screenshots of SGE showed ads still at the top of the page, and Google has indicated it will find ways to inject relevant ads even as the UI evolves. Bing Chat likewise started with some ads (including links within the chat response that were paid placements, and sidebar ads). Microsoft has explicitly said “ads can appear as inline placements in results” in Bing Chat and that as AI search grows, it expects to deliver more personalized, intent-driven ads through the chatbot marketingdive.com.

From the advertiser’s perspective, this means the search ad model is not disappearing but could become even more competitive. If fewer organic results are being clicked, businesses might feel a greater need to pay for guaranteed visibility via ads. For instance, if you run an e-commerce site and organic SEO is yielding fewer clicks because the AI gives an answer (perhaps even listing a product recommendation), you might invest more in paid search ads to get your link back in front of the user. Some analysts even predict the search ad market will grow thanks to AI, by unlocking new types of queries (people asking more conversational questions that the AI can then serve relevant ads against) axios.com. The Axios report cited eMarketer data suggesting the $350B global search ad industry could accelerate growth due to AI, not shrink axios.com. Why? AI could make search more engaging (more follow-up questions, etc.), giving more ad impression opportunities. It also enables multimodal searches – e.g. user drags an image to search or asks in natural language – where new ad formats might emerge.

However, advertisers do face challenges: measuring performance gets trickier in a world of fewer clicks. If an AI summary answers the query, a user might not click an ad immediately but could be influenced by the content. The classic “last-click” attribution model (where the value is assigned to the last link clicked before conversion) might undervalue the role of being present in the AI answer stage. This is similar to how featured snippets and voice answers have been – they influence without clicks. Advertisers and marketers will need to look at metrics like brand lift or consider multi-touch attribution (was the user exposed to our brand in an AI answer, then later came to our site via another channel?). Google might eventually provide new analytics for this, but currently it’s a gray area.

There’s also concern for cost-per-click (CPC) trends. If AI summaries reduce the supply of organic clicks, demand for the remaining ad clicks could push CPCs higher. Marketers with big budgets may pay a premium to ensure they’re visible in an AI-dominated result page (especially for high-intent queries). Smaller advertisers could be squeezed out if prices climb.

Interestingly, Microsoft’s approach hints at a partial solution: sharing ad revenue with content providers in AI results marketingdive.com marketingdive.com. For example, if Bing Chat shows an excerpt from Wired in its answer and places an ad next to it, Bing has floated giving Wired a cut of that ad revenue. If Google were pressured (via market or regulation) to adopt a similar model, it could compensate publishers indirectly for lost traffic. Advertisers might not see a difference, but it could keep the ecosystem somewhat balanced. As of Q3 2025, Google has not announced any such revenue share for SGE – it’s a point of contention.

SEO & Content Marketing Agencies: SEO practitioners are frantically adjusting playbooks. Some tasks remain the same (technical SEO, link-building, etc.), but there’s now the added element of optimizing for AI visibility. Agencies are advising clients to secure those coveted citation spots in AI answers. Part of this involves authority building: creating content and signals across the web that establish your site as an authoritative source. Marketers have been buzzing about the term “authority” – “I’ve heard the word authority more times than I can shake a stick at,” one marketing lead quipped digiday.com – because the belief is that AI algorithms will favor sources that seem consistently authoritative on a topic. This means businesses need to ensure their blogs, social media, and website content all align to reinforce expertise. For instance, if a company sells hiking gear, they want their site and blog to be known for hiking knowledge so that an AI answering “how to choose a hiking backpack” might cite them.

SEO firms are also monitoring how different queries behave. Some have found that not all query types get AI summaries – Google seems to be triggering SGE more for longer, question-like queries and fewer for one-word or navigational queries pewresearch.org pewresearch.org. Also, Google was tweaking placement: by June 2025, Google actually reduced how often the AI box appears at the very top (dropping from 98% to ~88% of the time on desktop), allowing the first organic result to appear above it more often digiday.com digiday.com. This was possibly a response to data or user feedback. For SEO, this means classic position #1 still matters because in a chunk of cases you might outrank the AI box, or at least appear immediately below it, which might salvage some clicks. SEO experts like Mark Traphagen suggested an “equilibrium” may be reached where Google shows AI for queries where users want an overview, but will refrain for queries where users tend to click results directly (Google can detect if people scroll past AI or ignore it) digiday.com digiday.com. If so, part of SEO strategy is understanding which keywords are likely to have AI results and focusing efforts accordingly. For example, a publisher might accept that very generic questions will have AI answers and not rank well, so they instead target more specific, long-tail queries or content that AI can’t easily summarize (like personal stories, unique data).

Affiliate Marketing: The affiliate model – where content publishers (like review sites, deal sites, influencers) drive traffic to e-commerce or services and earn commission – is especially vulnerable. Many affiliate sites rely on SEO to capture users searching “best X product” or “X vs Y comparison”. AI assistants are very good at answering these kinds of questions succinctly. Ask an AI “What’s the best budget DSLR camera?” and it might give you a top pick or a short list, possibly with direct links to retailers. That means a user bypasses the affiliate blog that spent effort writing a 2,000-word comparison (with affiliate links). As the Awin affiliate network noted, “Search’s AI revolution may be a threat to the affiliate longtail”, referring to the plethora of small affiliate sites that depend on Google rankings awin.com. In early SGE tests observed by SEO experts, some sites in sectors like recipes or product reviews saw organic traffic drops of up to 40% awin.com. If those sites’ revenue is mainly affiliate commissions, that drop is devastating.

What can affiliates do? The Awin report suggests affiliates continue serving their audience effectively and lean into Google’s quality guidelines (E-E-A-T) to stand a chance of being featured in AI results awin.com. Essentially, only the most authoritative and trustworthy affiliate content might survive – e.g. a well-known site like Wirecutter (owned by NYTimes) might still get cited or clicked because of its reputation, whereas dozens of smaller “top10gadgetsblog dot com” sites might lose out. Affiliates also might diversify traffic sources, similar to publishers: building email lists, using social media, and creating unique value beyond just aggregating info that AI can replicate. Some are exploring more first-person or experiential content – something AI can’t fake easily. For instance, a travel affiliate might focus on personal travelogues (with affiliate links embedded) which are harder for AI to summarize than a generic “10 best hotels” list.

Another angle: partnerships and integrations. We might see affiliate networks or platforms strike deals with AI providers such that the AI will include affiliate referral codes when it does provide links. For example, if ChatGPT recommends a product and offers a link to buy on Amazon, could that link include a referral for some trusted content partner? OpenAI hasn’t indicated this, but it’s conceivable as a way to give credit. Microsoft Bing, interestingly, has its own affiliate-like program via Microsoft Start partners (they share revenue on content that appears in some feed-based contexts) blogs.bing.com. Bing said it’s exploring showing a “rich caption” from licensed content alongside chat answers, with revenue share blogs.bing.com. If that model expands, maybe affiliate publishers could license their content to Bing or others to ensure they’re woven into the answer (and get paid for any sales that result).

Finally, advertisers who rely on affiliate channels might shift budgets. If affiliate sites send fewer customers, advertisers might pay more to other channels (like Google Ads, or direct partnerships, or retail media networks). For instance, an electronics brand that used to count on affiliate review sites for traffic might instead invest more in retail media (ads on Amazon, Best Buy, etc.) since users might jump straight from an AI answer to a retailer. This means the affiliate ecosystem’s loss could be the big retailers’ gain, consolidating power further.

In essence, search marketers are confronting an environment with less predictability. The old playbook of “optimize for top ranking, get clicks, get conversions” is being upended by a new intermediary – the AI answer – that might short-circuit the journey. Those who adapt by focusing on brand trust, diversifying traffic, and collaborating with the AI platforms stand the best chance to continue thriving. Those who don’t may see their traffic (and revenue) steadily siphoned away as the AI chatbots become the new gatekeepers of consumer attention.

Different Approaches, Shared Challenges: Platform Comparisons and Expert Opinions

It’s illuminating to compare how various platforms are handling AI summaries and how that impacts stakeholders:

• Google vs. Bing: While both giants now deploy AI answers, their approach in dealing with publishers diverges. Google took a more self-confident (some say combative) stance – launching SGE in Search Labs, rolling it out to users, and assuring publishers that all is well (“billions of clicks” are still flowing blog.google). Google emphasizes its data showing stable traffic and even claims it drives “more queries” and deeper engagement blog.google blog.google. In a polished blog post, Google’s Liz Reid gently dismissed third-party traffic decline reports as “flawed methodologies” or premature blog.google. The subtext: trust us, we’ve got this. Google’s bet is that by keeping users happy, long-term everyone (including publishers) will benefit from an expanded pie of search activity blog.google. They tout that AI will “empower us all to ask vastly more questions” and bring more engaged audiences to creators blog.google. Critics are skeptical – as Jason Kint said, publishers have “seen this movie before” whenever Google makes PR promises, only for the data to show otherwise digiday.com digiday.com. Google’s approach is certainly Google-centric: they are fine-tuning how often AI appears (e.g., pulling it back for some queries) to ensure user satisfaction, but not necessarily to save publisher traffic digiday.com. Any concessions to publishers so far (like snippet controls or the new Google-Extended tag to opt out of AI training digiday.com) have been minor. That could change if regulatory or reputational pressure grows. Microsoft’s Bing, on the other hand, being the underdog, made overtures to publishers from day one. Yusuf Mehdi wrote in March 2023 that Bing’s top goals included driving more traffic and revenue to publishers and doing so in a “collaborative fashion” blogs.bing.com. Microsoft integrated citations prominently and started convening partner meetings to gather feedback blogs.bing.com. They even floated specific new UI ideas (like the hover-over publisher link expansion) to directly boost clicks to sites blogs.bing.com. Furthermore, Microsoft signaled flexibility on the money front – e.g., “exploring placing ads in chat and sharing the ad revenue with partners” blogs.bing.com. This is a fundamentally different tone. And it likely stems from different incentives: Google already has the traffic and revenue and wants to guard it (while improving user experience to fend off competition from Bing/ChatGPT). Microsoft has less to lose from trying a more publisher-friendly model; in fact, it has upside if it can woo big content players to actively support Bing. It’s worth noting that in mid-2023, some publishers (like News Corp, owner of WSJ, and other media companies) were reportedly in talks with both Google and Microsoft about payments for content used in AI businessinsider.com. If Microsoft is more willing to cut checks or share revenue, it might gain favor – but Google’s dominance still gives it the upper hand (publishers can’t practically boycott Google Search, but they can collaborate with Bing as a supplement). In terms of impact, Google’s SGE clearly affects a broader swath of traffic simply due to volume. Bing’s AI might actually send more traffic proportionally because of how it’s designed with link prompts. For example, Bing Chat often suggests “I found this on [Site], would you like to read more?” and even opens a mini browser pane if the user clicks. That said, Bing’s share (hovering ~3–4% globally gs.statcounter.com) means even a perfect Bing Chat wouldn’t rival the absolute traffic Google can send or take away. One could argue Bing’s approach demonstrates that AI summaries and publisher referrals aren’t mutually exclusive – it’s possible to design for both user convenience and clicks out. Google’s design could evolve similarly (perhaps they will add a more visible “Explore more from [Source]” below AI answers if criticism mounts).
• OpenAI/ChatGPT vs. Search Engines: ChatGPT (and similar AI assistants like Anthropic’s Claude) represent a different paradigm – one that is not beholden to the web’s “attention economy” at all (at least currently). ChatGPT doesn’t need to keep you clicking on ads; OpenAI’s model is subscription and API-based. Thus, it has no inherent incentive to send you to publishers’ sites. The only incentive to cite sources or link is if users demand transparency or if not citing starts creating legal issues (like accusations of plagiarism or libel). So far, user demand for sources is mixed – some power users ask ChatGPT for sources or use plugins like VoxScript to get citations, but the average user might just accept the answer given. This free-floating AI is in some ways the purest threat to the old web order. It could become the new interface to information, period. A bleak scenario for publishers is if people increasingly skip search engines and just consult AI assistants for everything: “Hey ChatGPT, summarize the latest news on climate change.” ChatGPT would generate a nice summary (possibly drawn from several news articles), and the user feels informed – but none of those news sites got a visit. Some experts have indeed termed this possibility the “death of the web” as we know it, or at least of the open web where traffic flows to content creators. However, we’re not fully there yet. Even as of 2025, there are limits to ChatGPT’s capabilities that send people back to Google or websites: up-to-the-minute information, detailed data, diverse opinions, multimedia content, etc. Also, the AI models still make mistakes, so savvy users sometimes want to verify by clicking sources. Pew’s study did find that “users very rarely clicked on the sources cited” in AI summaries pewresearch.org pewresearch.org, but that was in Google SGE where the sources are a bit hidden. In ChatGPT, sources are usually not even given unless requested. This behavior may change as literacy about AI’s limits spreads. It’s possible a culture of verifying AI answers will lead more users to demand citations or to cross-check. If OpenAI were to integrate a default source list (as some interfaces built on it do), it might drive some traffic. But given OpenAI’s current setup, publishers likely won’t see meaningful traffic from ChatGPT unless it’s via some plug-in or integration that funnels users out. OpenAI did, at least, acknowledge content creators by starting to pay for some data (AP, etc.). And notably, OpenAI launched a web browsing mode that actually visits pages – this means, in theory, they might get into the business of sending referral traffic. For instance, if ChatGPT answers a user’s question by fetching a specific article during the chat, it might provide a link or the user may ask to see the full article. In practice, though, ChatGPT’s browsing was often used to just grab content and show it directly, not to encourage reading the actual page (though it would list the URL it visited). The dynamic with ChatGPT and publishers may eventually need a more formal détente – perhaps some “Content Provider Program” where publishers provide API access to their content for a fee or something, so that ChatGPT can reliably cite it. That might mirror how say, Google struck deals with news publishers for Google News Showcase (paying them to curate content).
• Perplexity and Smaller Platforms: These have positioned themselves almost as pro-publisher AI solutions – always citing and often limiting answers to entice clicking. But their scale is limited. Perplexity, for example, might drive a trickle of tech-savvy users to sites, but not in numbers that move the needle. That said, they serve as proof-of-concept that AI search can be done with respect for the open web. If Google and Bing face enough backlash, they could adopt more of those philosophies (indeed Google’s linking is more explicit than it was in early tests, after feedback).

One common thread across all platforms is that user behavior is shifting and everyone is trying to adapt. Even if Google and Bing differ in tone with publishers, both are seeing users spend more time interacting with AI on their sites. That’s arguably a win for them – keeping users in their ecosystem (zero-click benefits the search engine because the user still views ads or at least doesn’t leave to another site). In fact, Google’s revenue has not cratered; if anything, it continues to grow year-over-year. This implies that the loss of organic clicks hasn’t hurt Google’s bottom line – possibly because ads are still being clicked or because Google finds ways to monetize within the AI experience. As one tech analyst noted, Alphabet (Google’s parent) made ~$350B in 2024 largely from search advertising thefuturemedia.eu, and there’s little sign yet of that drastically falling. So from Google’s perspective, the shake-up might not be about revenue loss but about optics and ecosystem health. They don’t want to kill the golden goose (good content to index), so they need publishers to survive. That’s why we see Google repeatedly saying they care about the “health of the web ecosystem” and that Search’s “value exchange with the web remains strong.” blog.google blog.google. It’s a delicate balance: Google wants to evolve with AI to please users (and keep them away from rivals) while convincing content creators that this evolution isn’t going to bankrupt them.

Key expert opinions reflect this tug-of-war:

• On one side, optimists like Google’s spokespeople or some industry analysts argue this is a natural evolution. They believe AI summaries will expand search (people will search more and ask new questions), which in turn could expose them to even more content in the long run. They see it as similar to when search engines first appeared – yes, it disrupted older discovery methods, but it ultimately generated more traffic to more sites than the era of directories or offline info. Google frames AI as “one of the most expansionary moments for the web”, not a zero-sum game blog.google. Some independent analysts concur that search query volume is rising and that AI might actually stimulate new search habits (like multi-turn queries that dig deeper, potentially handing off to clicks when the user wants detail).
• On the other side, many publishers and some economists see a potential breaking of the internet’s traditional value loop. If content creators can’t get traffic or revenue, they can’t afford to create content, and the AI will eventually run out of quality material to summarize (or at least, quality will diminish). This is often illustrated by the phrase “feeding on its own exhaust” – if AI ends up mainly scraping AI-generated content because original content dwindles, information quality could degrade. “It isn’t obvious until it plays out, but we’re seeing the breakup of Google’s model,” said one publishing executive, warning that an AI-dominated search could undermine the very sources that make Google useful fipp.com. In the Digiday podcast, the hosts pondered whether we’re witnessing a “culling of publishers” and an information ecosystem where only large, established players or those with alternative revenue survive digiday.com. They noted non-news content (like how-tos, hobby blogs) might suffer more – which could lead to less free information on those topics in the future.
• Experts in SEO like Rand Fishkin have been vocal that Google’s shift to zero-click (even pre-AI) was already harming the open web. He suggests that Google’s own properties and now AI answers are siphoning off so much traffic that a new balance must be found phocuswire.com. His data showing roughly only one-third of searches resulting in website clicks is often cited as a wake-up call to marketers.
• Some marketing strategists see opportunity in the chaos: they argue brands that build trust and authority can still thrive. As one marketer said, it’s about ensuring “when these AI bots crawl the web, [our brand] shows up” consistently digiday.com. There’s an emphasis on going beyond SEO tricks to genuine thought leadership so that AI will inevitably quote you. This perhaps favors bigger companies and well-resourced content teams, creating a power-law dynamic (the top sites might get an even larger share of what clicks remain, while smaller ones fade out).
• Affiliate industry voices are cautiously watching. Some, like in that Awin report, essentially advise affiliates to keep doing what provides value and hope that the cream rises to the top in AI results. They also mention the irony that page 2 of Google was already “dead” for SEO, and now AI might effectively compress even page 1. So the consolidation was arguably happening; AI is just speeding it up awin.com awin.com.

Finally, a noteworthy voice: Dave Maney, a media entrepreneur, captured the inevitability many feel: “I don’t think this wave can be stopped or turned back.” digiday.com Generative AI usage is skyrocketing (120 million+ Americans by end of 2025, eMarketer says digiday.com). User expectations are shifting towards immediacy and convenience. In Maney’s view, the focus should be on adapting business models to this new reality, rather than futilely trying to preserve the old status quo.

In conclusion, AI summaries are indeed rewiring the traffic flows of the internet. The ~$350 billion search advertising machine is being rejiggered – not necessarily shrunk, but redistributed and redefined. Publishers are losing some direct traffic and with it, ad dollars, forcing them to innovate and fight for fair compensation. Search engines are walking a tightrope between user satisfaction and ecosystem sustainability. Advertisers and marketers are rewiring tactics to remain visible in an answer-first world. Each platform’s approach – Google’s integrated AI, Bing’s collaborative stance, OpenAI’s standalone model – offers a different balance of risks and benefits to publishers.

The rest of 2025 will likely see heightened negotiations: perhaps formal agreements where AI firms pay licensing fees or share revenue with content creators, and possibly regulatory moves especially in Europe and other regions concerned with Big Tech’s power. We’ll also see technological responses – maybe new tags or protocols that empower publishers to control AI usage of their content (much like robots.txt did for search engines decades ago). And on the user side, we’ll see if the novelty of AI summaries sustains and grows, or if users swing back toward wanting to verify things themselves on source websites.

For now, the genie’s out of the bottle. The way people search and discover content has fundamentally changed in a very short time. The “$350B shake-up” is ongoing: some players will adapt and thrive, finding new ways to capture value in the AI-driven traffic paradigm, while others who relied on old patterns may struggle or fade. As one industry saying goes, publishers have to skate to where the puck is going – and increasingly, that puck is controlled by AI-driven interfaces. The challenge and opportunity will be in forging a new web ecosystem where AI and publishers can coexist symbiotically, ensuring users get quick answers and the rich context and creativity that only the broader web can provide.

Sources:

• Google Search Liaison (Liz Reid), “AI in Search is driving more queries and higher quality clicks” – Google Keyword blog blog.google blog.google blog.google blog.google
• Pew Research Center, “Google users are less likely to click on links when an AI summary appears in the results” (July 22, 2025) – user behavior study pewresearch.org pewresearch.org
• Digiday (Jessica Davies), “Google AI Overviews linked to 25% drop in publisher referral traffic, new data shows” (Aug 15, 2025) – DCN publisher survey data and quotes digiday.com digiday.com digiday.com
• Digiday (Sara Guaglione), “In Graphic Detail: AI platforms are driving more traffic — but not enough to offset ‘zero-click’ search” (July 10, 2025) – Similarweb analytics on referral trends digiday.com digiday.com
• Digiday Podcast (Kimeko McCoy et al.), “As AI rewrites search, publishers look for a lifeline” (Sep 9, 2025) – discussion with Digiday editors digiday.com digiday.com digiday.com
• Pew Research Center, analysis of 68,000+ Google searches with/without AI summaries (2025) pewresearch.org pewresearch.org
• Awin, “Search’s AI revolution may be a threat to the affiliate longtail” in 10 Affiliate Industry Trends 2024 awin.com awin.com
• Axios (Sara Fischer), “AI set to upend search advertising model” (Sep 6, 2025) – search ad industry outlook axios.com
• Bing Official Blog (Yusuf Mehdi), “Driving more traffic and value to publishers from the new Bing” (Mar 29, 2023) blogs.bing.com blogs.bing.com blogs.bing.com
• MarketingDive (Peter Adams), “Microsoft assures publishers as ads ramp up in AI-powered Bing” (Mar 30, 2023) marketingdive.com marketingdive.com marketingdive.com
• Professional Publishers Association (UK) statements to CMA on Google SGE impact digiday.com digiday.com
• SparkToro (Rand Fishkin), “2024 Zero-Click Search Study” (July 1, 2024) – clickstream analysis of US/EU Google searches sparktoro.com
• Jason Kint (DCN) quotes via Digiday digiday.com
• Jess Sholtz quotes on SEO/branding via Digiday digiday.com digiday.com
• Dave Maney quote on generative AI adoption via Digiday digiday.com