I had an account last spring that was doing everything right. Budgets held. Exact-match structure held. Landing pages held. Yet over six weeks, impressions stayed flat, clicks slid 19%, and booked jobs fell faster.
The client assumed we had a bid problem. We did not. We had a page problem.
Google had started answering the query above our ads. A large AI-generated block now occupied the space where position one used to do its work. Sometimes our ad appeared inside that block. Sometimes it showed below it. Same keyword, same bid, very different click.
If you run paid search in 2026, this shift is your job to understand. You cannot target it directly, turn it off, or ask Google for a discount when it changes how your ad gets read. You still pay full price for the click.
An AI Overview is Google answering a search directly instead of sending the user to 10 blue links. It appears at the top of the page for informational and commercial queries, summarizes a few sources, and includes links a user can expand.
For paid search, the important point is simple: there is no AI Overview campaign type and no opt-in checkbox. If your Search, Shopping, or Performance Max ad is eligible for a query, Google can place it inside the answer block with a small Sponsored label or in the usual ad slots around it.
I used to tell clients that an ad either won the auction or it did not. I was wrong. Now, winning the auction is step one. Step two is where Google puts the winner on a page with two different neighborhoods:
Same keyword. Same Quality Score math. Different attention.
Google makes that placement decision at serve time, and advertisers do not get to bid differently for each location. That is the new constraint: you control relevance, not placement.
Here is the sequence. Google runs the normal auction, then decides where eligible winners live: above the Overview, inside it, or below it. There is no placement bid adjustment and no opt-out.
Existing Search, Shopping, Performance Max and App ads are eligible above or below the Overview. Text and Shopping ads from Search, Shopping and Performance Max can show within it.
For the inside slot, your ad must match both the query and the content of the answer itself. That is why Google keeps pushing broad match and keywordless targeting. It is not subtle. Google wants systems that can interpret more than the literal keyword you typed into the ad group three years ago.
The practical result is often fewer clicks per impression. Your impression count can look fine while the answer absorbs the visit you expected to buy.
Seer tracked 3,119 informational queries and found paid CTR on AI Overview queries fell from 19.7% to 6.34% between June 2024 and September 2025, a 68% drop. A 2026 update added some texture: paid CTR rose slightly from 14.6% to 16.2% when an Overview showed, while brands cited inside the answer earned 91% more paid clicks than brands left out.
Those numbers do not point to a single neat outcome. They point to a messier but more useful one: the result depends on whether you are part of the answer or merely sitting underneath it.
Practical takeaway: treat a stable impression trend with a falling CTR as a placement problem until you have ruled it out. The answer may be doing the job your ad used to do.
The first move is unglamorous: stop paying for questions Google can answer before your ad gets a fair read.
If the query is “what is” or “how do” with no buying language, the Overview can answer it, and your ad becomes wallpaper. I pull the search terms report, sort for high-impression terms where AI Overviews show, and look for informational bleed:
Those terms get negatives or move into a separate observation bucket with a capped bid. You are not conceding volume. You are conceding clicks that were never likely to become pipeline.
That distinction matters. Plenty of account managers still defend cheap informational traffic because it might have “assisted value,” which is what the deck calls full-funnel presence and I call paying to be ignored. In an answer-heavy result, the user may get the information, leave, and never reach the landing page that was supposed to assist anything.
The second move is to feed Google the material it can actually read. Google says it considers both the user query and the Overview content when serving ads inside it. Only text and Shopping ads from existing Search, Shopping, and Performance Max campaigns can appear there.
In practice, broad match, Performance Max, and AI Max for Search are the routes into the conversational layer. Exact and phrase still matter. They defend the brand terms and high-intent searches you must own. They just will not do much to get you inside an answer assembled around a longer, more conversational query.
Say you spend $20k a month on home services:
[emergency plumber near me].Practical takeaway: defend obvious intent with exact match. Expand into answer-shaped commercial intent with systems built to interpret it.
Search still does the heavy lifting on defense. When someone types your brand plus pricing or reviews, an exact-match Search ad can hold the slot above the Overview, where intent is obvious and the click is still yours to lose.
Shopping works differently. For a query such as “best cordless drill under $150,” Google can pull your product title, price, and image directly into the answer because the Overview needs something to show, not just something to say. Feed quality now determines visibility twice: once in the old Shopping carousel and again in the answer itself.
Performance Max appears inside Overviews more than most managers expect because the structure makes it suited to the job. PMax has no keywords to gate it. It can match the meaning of the query and the meaning of the answer using your feed, text assets, and audience signals together.
Google confirms that Search, Performance Max, Shopping and App campaigns are all eligible around Overviews. Only PMax brings all three asset types to that auction at once.
So the account structure is not complicated:
That is not a reinvention of paid search. It is a cleanup of the parts most accounts already neglect.
Google does not give you an Overview report. There is no segment for showed inside the answer versus showed below it, no placement column, and no useful auction breakdown.
Your Search terms report lumps those outcomes together. A blended CTR of 4.1% can hide two very different realities:
So I stopped trying to measure the box directly. Measure around it instead.
Pick 20 high-impression queries where you know the Overview shows. Tag them. Then track CTR, conversion rate, and cost per qualified lead week over week against a control set where it does not show. That split gives you, within two weeks, a useful answer to the question Google will not answer in the UI.
The split should also change how you bid. When the click moves lower in the journey, optimizing for the click fails. I moved two accounts last year from target CPA on form fills to offline imports tied to closed-won revenue and pipeline value. Both saw reported CPA rise 12% to 15% for three weeks before sales-accepted leads recovered.
The mechanism was simple. Smart Bidding had been buying cheap informational clicks that never closed. Once the conversion signal became revenue, it stopped chasing answer-box traffic and paid more for the smaller set of clicks below the answer that still bought.
Practical takeaway: if Overviews cover more than a third of your impressions, your conversion action needs to represent money, not merely a form submission.
Most of what Google calls a Google Ads AI tool will not solve the Overview problem. Automatically created assets reshuffle headlines. Smart Bidding squeezes the auction you already won. Useful sometimes, but neither changes whether you can match the query-and-answer combination that earns a place in the Overview.
The setting built for this shift is AI Max for Search. AI Max expands what your Search campaign can match to. Search term matching finds the longer conversational query, text customization rewrites the ad to fit that query, and final URL expansion selects a landing page that answers it.
Google reports that campaigns using the full set see an average of 7% more conversions or conversion value at a comparable CPA or ROAS versus search term matching alone. The mechanism matters more than the headline number: broader matching plus a relevant page is what gives the ad a chance to qualify inside the answer.
I run it as a controlled test, not a switch I flip because a Google rep used the phrase AI-powered growth 11 times in 20 minutes.
In two B2B accounts, that approach cut wasted informational spend by roughly a quarter because the system learned to stop paying for answer-box impressions that never produced a sales-accepted lead.
This will not work for everyone. If you spend under $10k a month or close deals from three keywords, skip it. You do not have enough conversion volume to teach the system, and broad expansion will simply buy noise faster. If you run broader intent with real conversion volume, this is how you stay citable without hiring someone to watch search queries at midnight. That is exactly the kind of continuous query mining groas runs on the paid search side.
If I inherited your account tomorrow, I would not rebuild it. I would triage it.
Stop paying for answers. Pay for intent.
That means one hour in the search terms report tagging queries where an Overview showed, one pass adding negatives for pure information seekers, and one check that your conversion action connects to qualified pipeline rather than raw form fills.
I used to defend keeping cheap informational clicks for assisted value. I was wrong. In an answer-box result, they may assist nothing. They just cost you $4 to learn that Google already gave the answer away for free.
Here is the first-pass checklist:
Google will put more answers above your ads before it puts fewer there. The managers who keep CPA in range will treat the Overview as inventory to qualify for, not traffic to chase.
Own brand terms with exact match. Stay citable on commercial questions with clean feeds and an AI Max controlled test. Bid to revenue instead of clicks.
Practical takeaway: audit the answer first, then spend. The click still matters. It just happens one screen lower than it used to.
