The editorial argues this move was inevitable because search advertising represents roughly 57% of Alphabet's revenue, and AI Mode directly cannibalizes the ten-blue-links SERP that those ads were built for. Google had to port the ad model into the new surface or watch its core business contract.
The editorial highlights that ads will be targeted not just to the initial query but to the full conversational thread preceding the ad slot, with the model itself deciding when commercial intent is high enough to interrupt. Combined with the GA release of keywordless AI Max for Search, this represents a fundamentally new ad surface — narrative rather than keyword-shaped.
Google's official announcement positions the change as a natural extension of Search and Shopping campaigns — existing advertisers don't need to take any action, as their inventory automatically becomes eligible to appear inside AI-generated answers. The team frames placements as appearing 'where they're most useful,' emphasizing utility rather than monetization pressure.
The submission drew 491 points and 418 comments on Hacker News, with the editorial characterizing the reaction as 'predictably grim' — reflecting widespread community skepticism that injecting ads into AI-generated conversational answers will preserve answer integrity. The high engagement signals that the developer audience views this as confirmation of long-feared enshittification of AI search.
At Google Marketing Live on May 21, Google confirmed what the ads industry has been bracing for since AI Overviews shipped: paid placements are coming inside AI Mode, the conversational search experience Google rolled into general availability earlier this spring. The announcement, posted to the Google Ads & Commerce blog and discussed in a 491-point Hacker News thread, frames the change as a natural extension of Search and Shopping campaigns — existing advertisers don't need to do anything; their inventory will simply become eligible to appear inside AI-generated answers.
The mechanic is simple to state and complicated to live with: ads will render contextually inside the AI response, targeted not just to the initial query but to the full conversational thread that precedes the ad slot. Google's product team described it as ads being placed "where they're most useful," which in practice means the model decides when commercial intent is high enough to warrant interrupting itself. Shopping ads, sponsored product carousels, and standard text ads are all in scope. The rollout starts in the US in English, with no firm timeline for other markets.
Google also previewed agentic ad workflows — campaigns that an AI agent can adjust in-flight based on conversational signals — and confirmed that AI Max for Search, the keywordless matching system launched last year, is now generally available. The combination is what matters: a query surface that's narrative rather than keyword-shaped, paired with a bidding system that no longer needs keywords to operate.
The Hacker News reaction was predictably grim, and predictably right about one thing: this was never optional. Search advertising is roughly 57% of Alphabet's revenue, and AI Mode cannibalizes the classic ten-blue-links SERP that those ads were architected for. Either Google ports the ad model into the new surface or it watches its core business contract while serving more expensive inference. There was never a version of this where AI Mode stayed ad-free.
What's genuinely new is the auction surface multiplication. A traditional search session generates one ad auction per query. A conversational session with five follow-ups generates five — and each subsequent auction has more context than the last, because the model carries the conversation history. That's a structural change in how attention is priced. The advertiser who wins the first turn isn't guaranteed the second, and the model's summary of your earlier turns becomes part of the targeting signal whether you typed it explicitly or not.
There's a quieter problem buried in the announcement: provenance and disclosure. Google says ads will be "clearly labeled," which is what they said about AI Overview citations, and which a chunk of the HN thread spent dunking on. When a synthesized paragraph blends organic facts with a sponsored recommendation, the label is doing a lot of work. The FTC's endorsement guidelines were written for a world where the ad and the editorial content were visually separable. Conversational AI breaks that assumption by design — the sponsored bit is grammatically embedded in the answer.
For publishers, this is the second shoe. AI Overviews already cut referral traffic for informational queries by double digits in multiple SimilarWeb and Similarweb-adjacent datasets over the past year. Now Google is putting its own ad inventory inside the surface that replaced your inventory, while still using your content as training and grounding data. The economics for any site that depended on "how to" or "best X for Y" traffic just got worse, not because rankings dropped but because the ranked result is now a synthesis with paid placements stitched in.
The community reaction also hit on a less-discussed angle: incentive alignment of the model itself. If ad revenue is a function of how often AI Mode surfaces commercial intent, there is now a measurable, optimizable reason for the model to drift toward commercially-shaped answers. Google will swear up and down that ranking and generation are walled off from monetization, and they probably are at the model level. But the product surface — when to show a shopping carousel, how prominently to feature a sponsored result, how to phrase the transition from answer to ad — those are all tunable knobs, and the gradient points one direction.
If you run paid search, three things change this quarter. First, your negative keyword lists are now partially obsolete in AI Mode contexts. Conversational queries don't have keywords in the same way; AI Max already broke the keyword-as-contract model, and conversational targeting finishes the job. Audit which of your campaigns are opted into broad AI matching and decide whether you trust Google's intent inference enough to leave the dial there.
Second, your landing page strategy needs a conversational fallback. Click-through from inside an AI response carries different intent than click-through from a ranked link. The user has already received a partial answer; they're clicking because they want depth, a price, or a purchase path. Pages built for the "someone googled this and is starting their research" persona will convert worse than pages that assume the user already knows what they want.
Third, if you're a publisher or a SaaS competing for organic traffic, assume informational query traffic from Google is on a multi-year decline and rebuild your acquisition model around channels Google doesn't intermediate. Newsletter, podcast, community, direct integrations, partnerships — the channels that look unglamorous on a deck are the ones that survive the surface migration. "SEO" as a job title is increasingly "how do I get cited by an LLM," and the answer to that is roughly the same as the answer to "how do I get cited by a human": be the canonical source, be structured, be cite-able.
The interesting test isn't whether ads work in AI Mode — they will, because the auction is the same primitive — but whether users tolerate them at the density Google needs. Classic SERPs are about 60% ad real estate above the fold for commercial queries; if AI Mode answers reach that ratio, the product becomes self-defeating. Watch for the first quarter where Google reports AI Mode-specific monetization numbers, and watch whether competing assistants — ChatGPT Search, Perplexity, Claude with web — follow with their own ad surfaces or hold the line on subscription-only. The answer to that question decides what conversational search looks like in 2027.
It’s good to know that Google search will essentially be useless for me moving forward.
I would have expected them to wait with ads until OpenAI starts first and users switch to Gemini. Google is probably the player that could afford to wait the longest with this and increase their market share that way.
> Now, if someone searches for an espresso machine, Gemini will pull up your most relevant products and instantly write a custom explainer highlighting why your product may be the right choice for them.This is like the essence of the evil of AI ads distilled down to one sentence. For an advertise
I wonder whose bright idea it was to label ads as 'helpful'. Do Google execs actually look for ads first when they google a question?
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.
The particularly worrying thing here is that they're now going to be gathering training data for a conversational model on _how to influence people effectively even when they already know they are being influenced_. Even more than RLFH already does. "We had to build the Torment Nexus so ou