Google Sells You an AI Mode Ad Next to a Rival's Pick

Share
Cover banner: Google Sells You an AI Mode Ad Next to a Rival's Pick

You are being offered a new place to spend money, and the pitch sounds reasonable. Buy the ad slot inside Google's AI answer, be there at the moment the buyer decides. What the pitch leaves out is that on most commercial searches the answer sitting beside your ad is pointing the buyer at somebody else, and Google will rent you that seat anyway. My position after ten years running Google Ads accounts and a practice built on getting businesses named inside AI answers: the ad buys you proximity, the citation buys you belief, and only the second one reliably turns into a booked call.

AI Mode is Google's chat-style search tab, where instead of ten blue links you get a written answer with a handful of sites named underneath as the sources it drew from. A citation is when your site is one of those named sources. The ad and the citation are separate purchases, and only one of them is for sale.

Google will rent you the seat next to a competitor's recommendation

SE Ranking looked at what actually appears in AI Mode for commercial searches, and Matt G. Southern wrote it up for Search Engine Journal on July 21, 2026. Google returned a text ad on 29% of the commercial keywords they analyzed. For most of those keywords, in their words, "the advertiser's domain was not among the sources the answer cited." Read that twice with your own money in mind. The business paying to appear was, in the majority of cases, not one of the businesses the answer named.

29%
Share of commercial keywords where Google AI Mode returned a text ad, and on most of them the advertiser's domain was not among the sources the answer cited.
SE Ranking analysis, reported by Matt G. Southern, Search Engine Journal, July 21, 2026.

Picture your own business for a second. You run a physiotherapy clinic in Leeds. Someone with sciatica opens AI Mode and asks which clinic to try. The answer writes a short paragraph about what to look for, then names three clinics it trusts, none of them yours. Your ad sits beside that paragraph, paid for, counted as an impression, reported as a win in a monthly deck. The buyer reads the recommendation, not the ad, and calls one of the three. You paid for the billboard outside the shop while the shopkeeper inside pointed every customer across the street.

The decision that follows is a sequence, not a budget. A paid impression beside an answer recommending someone else is proximity without endorsement, so earn the citation first and the paid impression lands next to your own name instead of arguing with a rival's.

Owners buy the slot because it looks like it is part of the answer

The confusion is honest. On a phone screen the ad and the answer occupy the same block of space, in the same typeface, inside the same result. It reads as one unit. It is two, built by two different systems with two different jobs: an auction that sells attention, and a model that decides who to trust.

Google's own framing is doing work here too. In its help documentation on ads in AI Overviews, Google states its case plainly.

“Google internal data shows that people have been finding the ads within AI Overviews helpful because they can quickly connect with relevant businesses, products and services to take the next step at the exact moment they need them.”
Google Ads Help, "About ads and AI Overviews." This is Google's own internal claim about its own product, not an independent finding.

I have no way to check that, and neither do you, because the data behind it has not been published. Google's line covers ads in AI Overviews, the AI summary on the regular results page; the independent count covers AI Mode, its chat-style sibling. What has been published is the SE Ranking count. So you have a vendor claim about helpfulness on one side and an outside analysis of who actually gets cited on the other. When those two disagree, spend against the one you can verify.

The pitchWhat the analysis found
Be present where buyers decideA text ad appeared on 29% of commercial keywords tested. The buyer decides from the sources the answer cites, and the advertiser's domain usually was not one of them.
People find these ads helpfulSourced to Google internal data in Google's own help article. No underlying numbers released, so it cannot be checked by an advertiser.
The ad connects you to the moment of needThe moment of need is answered by the written recommendation above the ad. Placement next to that answer is not inclusion in it.
Sources: SE Ranking analysis via Matt G. Southern, Search Engine Journal, July 21, 2026; Google Ads Help, "About ads and AI Overviews."

There is a second reason owners get this wrong, and it is the older habit. For fifteen years, buying the top of the page meant buying first position in the buyer's attention. Nothing above you was making an argument. Now something above you is making an argument, and sometimes the argument is for your competitor. The auction did not change. The page did.

Proximity, citation and recommendation are three different things you can buy or earn

Frederick Vallaeys, who co-founded Optmyzr after years at Google, describes what he calls the shortlist economy: AI systems increasingly hand a buyer three to five options rather than a page of results. That framing explains the whole problem in one move. If the machine is building a shortlist, the only question that matters is whether you are on it. Everything else is decoration around somebody else's shortlist.

What you getHow you get itWhat the buyer takes from it
Paid impression in AI ModeBought at auction, labeled as an ad, available today.You exist and you paid to be seen. No opinion attached.
CitationEarned. The answer pulls a fact or a line from your page and names you as a source.The system found you worth quoting. Credible, but still not a nudge to call you.
RecommendationEarned. The answer names you as the one to use.A decision made on the buyer's behalf. This is what produces the call.
Framing draws on the shortlist concept described by Frederick Vallaeys, Search Engine Journal, July 14, 2026.

The middle row is the one most businesses misread, and I wrote a whole piece on it because clients kept celebrating the wrong milestone. Being cited by Google is not the same as being recommended by it, and the distance between those two states is where most AI visibility budgets quietly die. Add a paid slot on top of an uncited page and you have stacked a purchase on a gap.

The order is check, earn, then buy, and most accounts run it backwards

Nobody should read this and cancel their search advertising. Paid search still works, and for some businesses it is the fastest revenue in the account. What should change is the sequence, because a slot bought before a citation exists is spending against your own absence.

THE SEQUENCE I WOULD RUN BEFORE BUYING THE SLOT
1List your ten money questions. Not keywords. The ten questions a real customer asks out loud before they buy from you, in their words, including the ones with your city in them.
2Ask each one in AI Mode and write down who it cites. Take a screenshot. You are looking for two things: whether you appear at all, and which competitors keep reappearing.
3Open the pages it does cite and find what they answer that you do not. Usually it is specifics: real prices, real timelines, a named person, a plain answer to the awkward question you avoid on your site.
4Rewrite your own pages so they can be quoted. One question per section, the answer in the first two sentences, your name and your numbers in the text rather than in a picture.
5Then price the slot, question by question. Where the answer already names you, a paid impression reinforces your own name. Where it names three rivals and not you, treat that spend as a test with a stop date, not a line item.
Sequence based on my own client work. The finding it responds to: SE Ranking via Search Engine Journal, July 21, 2026.

Step two is where most owners stop, because they check once, see nothing, and conclude they are invisible. A single check tells you almost nothing, since these systems answer the same question differently across sessions. Run it properly instead, the way I lay out in this guide to checking whether AI actually recommends your business, then you have a baseline you can argue with.

If you have an agency, you do not need to do any of this yourself. You need to ask one question and listen carefully to the shape of the answer. Ask: for our top ten commercial queries, does AI Mode currently cite us, and what is the plan to get cited on the ones where it does not? An agency that has looked will answer with a list. An agency that has not will answer with a paragraph about visibility.

FIVE QUESTIONS TO PUT TO WHOEVER RUNS YOUR ADS
Show me the answer. For our five biggest queries, send me a screenshot of the AI Mode answer and highlight who it cites.
Separate the reports. Ask, in writing, what spend the platform can and cannot break out by AI surface, then track booked jobs against total search spend for the test window, not clicks.
Name the rivals. Which three competitors keep getting cited on our queries, and what are they publishing that we are not.
Set the stop date. If the AI Mode test does not produce booked work in ninety days, what happens to that budget.
Prove the tracking works. Show me one booked call traced end to end, from the query to the calendar, before we scale anything.
Ask these in writing. The answers are the audit.

Impressions inside an AI answer are the easiest vanity metric ever invented

Every new placement arrives with a friendly number attached, and this one is friendlier than most, because "we appeared inside the AI answer" sounds like arrival. It is not arrival. It is rent. The only numbers worth putting in front of an owner are the ones that survive contact with the bank account.

Track three things monthly. First, booked calls and their cost, not clicks and their cost. WordStream and LocaliQ put the cross-industry average cost per click at $5.42 and average cost per lead at $66.69 in their 2026 benchmarks, updated May 19, 2026, which means a modest test can burn a few thousand dollars before anybody has a strong opinion about it. Second, your citation count on those ten money questions, checked the same way every month so the trend is real. Third, whether the answer that cites you also recommends you, because those move separately.

$66.69
What the average Google Ads lead costs across industries. Judge any AI Mode test by leads at this price, not by impressions.
WordStream/LocaliQ 2026 benchmarks, updated May 19, 2026. Average CPC across industries: $5.42.

The trap on the paid side is the same one I find in almost every account I audit, dressed in new clothes. Money flows to whatever reports well, and the new placement always reports well early. I have written before about the spend that leaks out of a Google Ads account while looking fine in the dashboard, and an AI Mode line item with impressions climbing and no traced bookings is exactly that leak with a fresh label on it.

There is one more reason the click number is losing its meaning. SparkToro's 2026 analysis found that fewer than one in three Google searches now sends a click to any website at all. If most searches end inside the answer, then the answer is where the buying decision is made, and the metric that matters is whether your name is inside it. I learned this before AI Mode existed: I filled a London ADHD clinic's calendar for three straight months, to the point where they hired more specialists and outsourced the overflow, and the thing that moved was never impression volume. It was being the obvious choice at the moment somebody decided.

Frequently Asked Questions

Are Google AI Mode ads worth it for a small business?

They are worth testing only after you know whether AI Mode already cites you for your main commercial questions. SE Ranking found ads on 29% of the commercial keywords they analyzed, and on most of those the advertiser was not among the sources the answer cited, so you can pay to sit beside an answer steering the buyer elsewhere. If the answer already names you, the ad reinforces your own name and is worth a controlled test. If it names three competitors and not you, spend the first budget on getting cited, then revisit the slot with a stop date.

Can I pay Google to be one of the sources its AI answer recommends?

No. The ad slot and the cited sources are two different systems. The auction sells you a labeled placement next to the answer, while the citations are chosen by the model from pages it considers worth quoting. That is why the sequence matters: earn the citation through what you publish, then decide separately whether the paid impression is worth buying on that query.

How do I know if AI Mode is showing an ad on my keywords?

Ask your ten most commercial customer questions in AI Mode yourself and look at what appears above and below the written answer. Note whether a labeled ad shows, whose ad it is, and which sites are named as sources. Repeat the same ten questions monthly, because these answers vary between sessions and a single check will mislead you. Screenshots of that pattern are more useful in a budget conversation than any report your ad platform will generate.

Search advertising taught a whole generation of owners that the top of the page is the best place to be. That was true when nothing on the page had an opinion. The page has an opinion now, it prints that opinion above your ad, and no amount of bidding changes what it says about you.

If you want to know what Google's AI actually says about your business before you fund another placement next to it, I will run your ten money questions and show you who is being named instead of you. You can book an AI visibility call here and we will look at the real answers together.

Read more

Free, No Commitment

Find out exactly where your AI visibility is leaking. In 30 minutes.

No pitch. No fluff. A straight diagnostic on your specific situation and the single highest-leverage fix to make right now.