How does AI Search affect media buying on the open web?

AI Overviews and chatbots reduce the click-through rate from search to publisher sites, shrinking the addressable audience behind open web display inventory. Index Exchange data from March 2026 shows 69% of publishers on its platform experienced year-over-year declines in ad opportunities in 2025, with an average drop of 14%. Forrester projected in late 2025 that open web display investment could contract by 30% as audiences migrate to AI-answered surfaces.

What is the difference between AI Search advertising and traditional paid search?

Traditional paid search (SEA) buys a ranked placement on a results page; the user sees the ad alongside organic results and clicks to a landing page. AI Search advertising — or Generative Engine Advertising (GEA) — places brand visibility inside the synthesized answer itself, in formats native to the conversational interface. The unit shifts from the click to the citation, and the publisher whose content was used to generate the answer becomes part of the value chain.

Should media buyers cut open web budgets now?

Not as a blanket move. The risk is not that open web inventory disappears but that it concentrates, becoming scarcer and more expensive while delivering smaller reach. Cutting blindly into walled gardens accelerates the platform dependency problem. The correct response is to model the contraction, hold quality open web positions, and begin allocating a test budget to emerging AI Search advertising inventory — treating AI visibility as a separate discipline with its own KPIs.

Why are publisher licensing deals with AI companies not solving the revenue problem?

Licensing deals compensate publishers for content used in AI training, but TollBit's 2025 data shows click-through rates from AI applications fell for publishers both with and without licensing deals. The structural problem is that readers no longer need to visit the publisher to get the answer. A payment rail that compensates publishers for being cited in AI-generated answers — not just for training data — is the missing piece.

What is SPUR and why does it matter for advertisers?

SPUR — Standards for Publisher Usage Rights — is a coalition launched in February 2026 by the BBC, Financial Times, Guardian, Sky News, and Telegraph to establish shared technical standards and licensing frameworks for AI use of journalistic content. For advertisers, it matters because it represents the supply side organising to define how premium content flows into AI answers — which will ultimately shape which publishers appear in AI-generated responses, and at what cost.

What is Generative Engine Advertising (GEA)?

Generative Engine Advertising is the emerging category of paid placement inside AI-generated answers — distinct from SEO (organic search ranking) and SEA (paid search on a results page). GEA places brands inside the synthesised response that an AI Search engine returns to a user query, with the publisher whose content informed that response forming part of the value chain. It is the advertising model built for an internet navigated by AI agents rather than by blue-link clicks.

Buy-Side AI Risk: Open Internet Inventory Shrinks in 2026

AI chatbots aren't just draining publisher traffic — they're quietly eroding the open internet's ad inventory. Media buyers who haven't repriced this shift yet are running out of time.

The open internet's traffic crisis has been framed, almost exclusively, as a publisher problem. That framing is wrong — or rather, it is now dangerously incomplete. AI Overviews and chatbots are not just siphoning readers from news sites; they are destroying the addressable audience that media buyers assume sits behind open web inventory. The buy side has a structural risk on its hands, and most media plans still don't reflect it.

Google search referrals to publishers fell 33% globally in the year to November 2025, and 38% in the US, according to Chartbeat data cited in the Reuters Institute's January 2026 report — meaning the audience behind open web display inventory is shrinking faster than media plans are being adjusted.
69% of publishers on the Index Exchange platform reported year-over-year declines in ad opportunities throughout 2025, with an average drop of 14%, per an Index Exchange analysis published March 2026.
Forrester projected in late 2025 that advertisers could cut open web display investment by 30% in 2026, redirecting budgets to CTV and paid social as addressable audiences on publisher sites contract.
As premium open web inventory becomes scarcer and easier to identify, tighter supply will shift negotiating power — driving up CPMs and increasing dependency on walled gardens.
The emerging parallel economy — AI-generated answers, citation-based brand visibility, publisher monetization through content licensing — is real and accelerating, but it currently lacks the ad infrastructure to connect both sides.

What the open internet's ad economy was built on

The open web's advertising economy is built on a closed loop: publishers produce content, search engines distribute it, readers arrive, and advertisers reach them. Every element of that loop depends on one thing — the click. Click-through from search to publisher is the mechanism that creates an addressable, monetizable audience. Remove the click and the loop breaks: no readers, no inventory, no ad market outside the walled gardens.

For twenty-five years that loop held. Google sent traffic; publishers built business models around it; media buyers priced it; agencies planned it. The CPC, the programmatic auction, the viewability standard — all of it assumes a reader who arrived via a link. The agentic web is quietly terminating that assumption.

Click-based media buying versus citation-based brand presence

In the search economy, brands competed for the ranked position above the fold, then paid per click. Visibility was a function of bid, quality score, and relevance — a system media buyers understood, planned, and measured. In the agentic economy, the answer engine synthesizes a response and cites the sources it used. The brand that surfaces in that answer does not buy a placement in a traditional sense; it earns a citation because its content was trusted enough to be ingested.

The transition is already visible in the data. TollBit's Q4 2025 State of the Bots report found that click-through rates from AI applications to publisher sites dropped from 0.8% in Q2 2025 to 0.27% in Q4 — a near-threefold collapse in a single year. Training scrapes fell 15% over the same period, while RAG (retrieval-augmented generation) bots rose 33% and AI search indexers rose 59%. The bot traffic is no longer primarily about building models; it is about answering user queries in real time, without redirecting those users anywhere.

What The Current reported, and what the broader data confirms

Zac Wang's March 2026 analysis in The Current is one of the cleaner articulations of a buy-side risk that the industry has been slow to price. Chris Milano, global VP of supply at Assembly Global, frames the stakes directly: as AI chatbots reduce direct visits to publishers, fewer ad opportunities exist. Michael Beuth, managing director at Mediaplus Group, adds the concentration angle: as premium open web environments become identifiable but scarce, incremental platform spend reaches diminishing returns.

The article cites a 55% jump in national news AI scraping in Q4 2025, sourced to TollBit. That figure is directionally correct for national news as a category, though TollBit's broader finding is a 20% overall scraping increase from Q3 to Q4 — with national news as one of the steepest verticals. The directional read holds: high-trust news content is being scraped faster, not slower, and the traffic that would have monetized it is not returning.

Why premium inventory scarcity is not a safe harbour for media buyers

The case for concentrating on surviving premium environments

A credible counter-argument — voiced by analysts including Evelyn Mitchell-Wolf at Forrester — is that as low-quality, commoditized traffic disappears, what remains is genuinely premium: high-attention, trusted environments where advertising impact is harder to replicate. The argument follows that media buyers who concentrate spend in proven premium inventory will achieve better outcomes with a smaller pool. Beuth's own data suggests this is already happening: platforms like Reddit and Roblox are attracting AI crawler attention, creating new surface areas for advertising alongside the surviving legacy premium publishers.

Why scarcity without a new inventory layer creates a dependency trap

Scarcity is a temporary market condition, not a strategy. When addressable audiences on the open web contract, budgets do not disappear — they migrate, typically into the platforms that can guarantee reach. That is exactly the dependency dynamic Beuth warned against in The Current: concentrating spend in platforms past their marginal utility does not produce better outcomes, it produces higher prices and reduced negotiation leverage. The real question is not where to concentrate remaining open web spend, but whether a net-new inventory layer is forming inside AI-generated answers. The evidence says it is — and it is forming without the publisher and advertiser side being properly positioned to capture it.

What AI Search means for media buyers, agencies, and publishers today

For CMOs, media buyers, and agencies: reprice the open web before the market does it for you

The structural shift is already priced into publisher traffic data; it is not yet priced into most media plans. Three actions follow. First, audit how much of your funnel's reach currently depends on open web display inventory and model what a 14–30% contraction in ad opportunities means for your 2026 and 2027 plans. Second, begin separating AI Search visibility as a line item — brands cited in AI Overviews saw paid CTR run 91% higher than uncited brands in Q3 2025, per Seer data, which suggests that citation is already a media performance signal, not just an SEO signal. Third, resist the reflex to absorb budget cuts by concentrating into the walled gardens — that path leads precisely to the leverage problem Beuth identified.

For publishers: the licensing deal is a floor, not a business model

The formation of SPUR — the Standards for Publisher Usage Rights coalition, launched in February 2026 by the BBC, Financial Times, Guardian, Sky News, and Telegraph — signals that major publishers understand they cannot negotiate individually against AI platforms at scale. SPUR's mandate to create shared technical standards and responsible licensing frameworks is the right infrastructure play. But licensing deals alone will not restore publisher economics. TollBit's data is explicit: click-through rates from AI applications declined for publishers with and without AI licensing deals in 2025. The revenue model that fits a citation-based web is one where publishers are inside the AI ad value chain — not just compensated for training data upstream.

Three signals pointing to where this lands

Three signals are already moving. Google launched shopping ads inside its AI Mode in early 2026, with sponsored placements appearing inside AI-generated responses for product queries — the first major test of native advertising inside an AI answer environment. Microsoft's Advertising launched AI Citations in limited preview in April 2026, giving brands visibility into how their content surfaces inside AI-generated answers. And Perplexity — which had been running sponsored follow-up questions before pausing new advertising deals in October 2025 to reassess measurement — signals that the category is real but the pricing and measurement infrastructure is still forming. The UK's CMA is separately requiring Google to allow publishers to opt out of AI Overviews without losing organic search visibility, a regulatory signal that the content supply chain for AI answers is now a competition matter, not just a copyright matter. The category is converging from multiple directions simultaneously.

Conclusion

Hold on to this: the buy side's AI blind spot is not a media quality problem — it is a structural inventory problem. As AI-generated answers replace the click-driven open web loop, the addressable audience behind open web display contracts, premium inventory concentrates and reprices, and the platform dependency trap tightens. The media plan that doesn't reflect this reality in 2026 will price it in painfully by 2027.

Smalk AI is built for exactly this inflection: an AI Search ad network that connects brands needing visibility inside AI-generated answers with the media sources whose content powers those answers — placing native ads for AI agents while opening a new revenue stream for the publishers whose traffic is being replaced by citations. What to watch next: which platform or standards body — SPUR, IAB Tech Lab, or a consortium of DSPs — sets the first widely adopted pricing benchmark for sponsored citations inside AI answers. That's the moment Generative Engine Advertising moves from experiment to line item.

The Buy Side's AI Blind Spot Is Becoming a Structural Risk