Skip to content
Skip to content
Paid Search

Google AI Mode ads arrive as a quiet text link, and regulated brands should read the label

On 14 September 2026 an SEO watcher posted a screenshot of a text link inside Google AI Mode with a small Sponsored label above it, and every fox who reads labels for a living sat up straight.

Summarise with
Set as preferred source on Google
18 min read
Quick answerGoogle AI mode ads are being tested as text links under a Sponsored label, spotted on 14 September 2026, not announced by Google. Regulated advertisers should audit policy-sensitive copy and landing pages before any rollout.
Section 01

What the google ai mode ads test actually shows

A startled fox in a birch thicket stops before a signpost with one brighter board, an image for google ai mode ads
Read the label before you read the link.

The fox does not trust a trail because it looks tidy. It reads the scent. Here is the scent as of 20 September 2026: on 14 September, SEO tracker Sachin Patel posted on X that Google was testing a new ad format inside AI Mode, and later that day added that the test also showed on desktop. The next day, Search Engine Roundtable's Barry Schwartz described the detail that matters. These google ai mode ads look like ordinary AI Mode text links, anchor text and all, set apart only by a Sponsored label above the AI-generated response.

Two cautions belong in the first paragraph, not the footnotes. First, this is a test spotted by an observer and reported by a trade publication. Google has not announced these google ai mode ads, and we found no Google page describing them as of 20 September 2026. Second, Schwartz noted that Google showed plain text anchor links in AI Mode responses in April, without ads, so what changed is that the same shape now carries a Sponsored tag. Nothing in the reporting says how many people see it, in which countries, or on which queries.

Schwartz was blunt about the commercial odds. He wrote that the units "look way to organic", and that he doubts their click-through rate will match the normal format, because people tend not to click links inside AI responses. Treat that as one seasoned watcher's forecast, not a measurement. No one outside Google has published a click-through figure for this unit, and no one inside Google has said one exists.

Why should a regulated advertiser care about google ai mode ads, which are nothing grander than a text link? Because a link woven into an answer is where your ad copy stops being a self-contained unit. The headline compliance signed off now sits beside sentences you did not write, generated at the moment of the query. The table below sets out what each surface documents today, and the empty cells are the story.

The two surfaces with paperwork both keep gambling out; the AI Mode text-link test, the one that prompted this piece, has published neither a category rule nor a reporting rule.
QuestionGoogle AI Mode text-link testGoogle ads in AI OverviewsOpenAI ChatGPT ads
StatusSpotted test, 14 September 2026; no Google page foundLive in 12 countries, per Google Ads HelpLive; Sponsored Agents piloted with select US advertisers
GamblingNo published position foundNot shownDisallowed, casino lodging and travel excepted
FinanceNo published position foundNot shownUS only, case by case, manual review
HealthcareNo published position foundNot shownUS only, case by case; mental health chats excluded
ReportingNo published position foundNo segmented reporting; Top AdsClick windows 7, 14, 30 days; view 0 or 1
  • StatusSpotted test, 14 September 2026; no Google page foundLive in 12 countries, per Google Ads Help Live; Sponsored Agents piloted with select US advertisers
  • GamblingNo published position foundNot shown Disallowed, casino lodging and travel excepted
  • FinanceNo published position foundNot shown US only, case by case, manual review
  • HealthcareNo published position foundNot shown US only, case by case; mental health chats excluded
  • ReportingNo published position foundNo segmented reporting; Top Ads Click windows 7, 14, 30 days; view 0 or 1

Read every "no published position found" as a question, not a permission. Absence of a rule is not a licence, and a quiet hedgerow is not the same as a safe one. The rest of this piece is about what the gaps in google ai mode ads paperwork mean for teams whose ad copy carries legal weight.

Section 02

What Google itself says about google ads in ai mode

Set the observer aside and read Google's own paperwork on google ai mode ads. On 20 May 2026, Google's Marketing Live announcement introduced conversational discovery ads, which answer a person's specific question with creative Gemini builds for that query, and Highlighted Answers, which let a relevant ad appear inside a list of AI Mode recommendations. Google said both formats will stay clearly labelled Sponsored, and that it will keep testing them.

Google's own Ads in AI Mode page says the same testing is under way, with the new formats sitting inside AI-generated responses, labelled as ads, and powered by AI Max and Performance Max. That eligibility detail matters if you have not yet read our audit of the AI Max migration, or our note on how Google let the tight keywords in in AI Mode tests. The trail into google ai mode ads runs through campaign types Google is moving accounts onto, whether you chose the move or not.

The sharpest primary evidence sits on a neighbouring surface. Google Ads Help, in About ads and AI Overviews, says Google does not currently show ads in AI Overviews for sensitive verticals like adult, alcohol, gambling, finance, healthcare and politics. It also says advertisers cannot opt out of AI Overview placements, that Google Ads offers no segmented reporting for ads shown there, and that placement is limited to English in twelve countries, the US among them.

That page is about AI Overviews, not AI Mode. We found no Google page that gives the same category rule for AI Mode, and we will not pretend the AI Overviews rule stretches neatly over it. It is a strong scent with no guarantee attached. The prudent reading is that Google's habit is to keep the money-and-health thicket out of new generative surfaces first and let it in later, if at all.

Notice what is missing from that list: any number describing how often the text-link version of google ai mode ads appears, how it performs, or who sees it. Google's own Search Central guidance is also plain that AI Mode traffic counts inside ordinary Search Console web reporting and that no special markup or extra files are needed to appear in these features. That is guidance for organic visibility, not for paid, but it settles one thing: there is no secret handshake, for ads or for organic. Do not buy one from anyone who claims otherwise.

Section 03

Which regulated advertisers can run ai mode ads and their cousins today

A rulebook with a monthly pulse
Line chart of cumulative published versions of OpenAI's ad policy, rising from one in March 2026 to seven in September 202686420MarAprMayJunJulAugSepPublished versions: 1Published versions: 2Published versions: 3Published versions: 3Published versions: 4Published versions: 6Published versions: 7
Published versions
Line chart of cumulative published versions of OpenAI's ad policy, rising from one in March 2026 to seven in September 2026
ItemValue
Published versions1
Published versions2
Published versions3
Published versions3
Published versions4
Published versions6
Published versions7
OpenAI's ad policy reached its seventh published version in September 2026, seven versions in seven months, so any category answer you cache goes stale fast. Source: the changelog on OpenAI's ad policies page.

That chart is a warning about your calendar. A rulebook that moves nearly monthly is a moving quarry, and a regulated team that checks it once a quarter is tracking last season's prints. The current version, OpenAI's ad policies page, was updated on 10 September 2026 and clarified that OpenAI may decline ads where they conflict with its advertising principles or business interests.

What does that page allow? Its placement rules keep ads away from conversations touching regulated goods (gambling and tobacco are the examples given), political content, and mental and personal health. Its content rules disallow gambling promotions outright, with casino hotels and travel as the narrow exception. Financial services and health services are restricted: in the US, OpenAI says it may approve named product types from approved advertisers case by case, and may ask for proof of licensure. Outside the US, financial and health ads are generally prohibited. The bars below count the product types the page actually names.

Product types OpenAI names as possible in the US
Financial services
10 types
Health services
9 types
Gambling
none, disallowed
Alcohol and tobacco
none, disallowed
OpenAI names ten financial and nine health product types it may approve case by case in the US, and none for gambling or alcohol.

Google's per-product rules are older and more granular, and they attach to the advertiser and the format today. Google's gambling rules require certification and approved countries, and a landing page that displays responsible gambling information. Its financial services rules require the physical address and all associated fees, clearly and immediately visible without a click or a hover. Its healthcare rules means that in the US you must be certified by Google to advertise health and medical insurance coverage. And its alcohol rules already list formats such as Discovery ads and dynamic search ads as ones that do not allow alcohol at all.

Whether google ai mode ads inherit those format lists is exactly the question Google has not answered in any page we could find. Until it does, the honest position for a casino, a lender or a clinic is to assume the strictest reading applies and that an answer engine might one day quote your ad beside a sentence a regulator would dislike.

One more scent for the moonlit margins: disclosure. The FTC's native advertising guide says an ad is deceptive if it is not readily identifiable as an ad, and advises placing the disclosure in front of or above the headline. A Sponsored label above the response is at least on the right side of that advice, though the guide was written for editorial-style ads, we are not lawyers, and none of this is legal advice.

Section 04

What OpenAI's sponsored agents pilot shows about where ai mode ads may go

Now the contrast with google ai mode ads. On 16 September 2026, OpenAI said in its own post that it is testing Sponsored Agents, which let people start a conversation with a business-sponsored agent after clicking an ad in ChatGPT. OpenAI describes the exchange as a clearly labelled conversation, distinct from ChatGPT's independent answers, and says it is being tested with select advertisers in the United States. Search Engine Roundtable covered the same pilot the same week.

Set the two side by side. Google's text link is a click away from your landing page. OpenAI's pilot is a conversation about your business, spoken by an agent under your sponsorship. For a regulated advertiser that agent is a second mouth, and every mouth needs a compliance sign-off. The pages we read do not say how the content of a sponsored conversation is reviewed, so ask before you enrol, and ask in writing.

A label tells the reader it is an ad. Only your compliance file can say it is the right ad.
folkfox, on ads that live inside answers

Measurement is where the two paths visibly part. OpenAI's measurement page lets advertisers pick a 7, 14 or 30-day click window and a view window of 0 or 1 day, applies to reporting only, and uses last-touch logic rather than multi-touch. OpenAI's event quality page says event quality refreshes daily over seven complete calendar days, and that its score does not predict campaign performance; Search Engine Roundtable described it as a 1 to 10 score. Set that against Google Ads Help, which says there is no segmented reporting for ads in AI Overviews, and against silence about the text-link test.

A fair reader will say OpenAI's tooling is still a pilot's tooling, and they would be right: reporting windows are not results. Neither company has published performance evidence for these units that we could find. The honest way to talk about scale is the way we did when we compared three studies of ChatGPT ad density: measured ad load swung enormously depending on the denominator, which is why we attribute every figure and assert no scale of our own.

Our view, plainly: the text-link test behind google ai mode ads is the least glamorous thing Google is trying and possibly the most instructive. It makes an ad look like an answer, and that is precisely where regulated copy is most exposed. We expect finance, gambling and health brands to be last in and rightly cautious, and we expect click-through to be lower than classic text ads, echoing Schwartz. That last part is opinion, not data. Do not build a forecast on a format nobody has measured.

Section 05

What a paid search team should do on Monday about google ai mode ads

The Monday checklist, in order
Audit sensitive assets

List every headline, description, sitelink and promotion that carries a claim: a rate, an odds line, a bonus, an outcome, a risk statement. Mark the ones compliance would not want quoted beside unseen text.

Make landing pages self-contained

Put the licence, risk warning, fees and responsible-play or health disclosures on the page itself, in plain sentences, so nothing depends on the ad wording to stay compliant.

Pre-decide exclusions

Agree now which query themes you would pull from any AI answer placement, such as advice-seeking finance and health queries, and use the controls you already have, negative keywords and campaign structure, to enforce them.

Reset measurement expectations

Capture a pre-period baseline, annotate the account when anything changes, and do not plan around the click-through of classic text ads.

Classify test versus rollout

Ask whether a Google Ads Help page, an announcement date, a policy line and a reporting column exist. If none do, file it as a test and revisit weekly.

Start with the audit, because it is the only step that pays off even if this test disappears next week. A fox that has mapped its den does not panic when the brush moves. Regulated teams already know which of their lines lean hardest on qualifiers, so the exercise is quick: pull the top spending campaigns, read the copy as if the second sentence were written by a stranger, and flag anything that would mislead without its caveat.

The second step is the one most often skipped. If an answer can quote your ad without your disclaimers, the landing page behind google ai mode ads must carry the full truth on its own. That is a job for content and compliance together, and it is also good practice for organic visibility; our iGaming marketing and FinTech marketing pages cover how we run it for regulated operators, with healthcare marketing for the health side.

Exclusions and measurement are two halves of one habit: decide before you are surprised. Google Ads Help says advertisers cannot opt out of AI Overview placement, so do not assume a switch exists for google ai mode ads. Instead, note which reports would show a sudden shift, and keep a weekly diary. Our note on proving the signal covers the baseline discipline, and the AI Max migration audit covers the account-side changes already landing.

The fifth step is the one that protects your sanity. Tests come and go, and screenshots are seductive; Google itself said in May 2026 that it would keep testing these formats. A rollout leaves paper: a help page, a policy line, a reporting column, a date. A test leaves a screenshot and a timeline. Until the paper appears, treat google ai mode ads as something to watch and prepare for, not something to budget for. If you want a second pair of eyes on a regulated account, our paid search team runs exactly this audit, and the pricing page shows what it costs.

Questions

Frequently asked questions

What are google ai mode ads?

Google AI mode ads are paid placements shown inside Google's AI Mode. Google said on 20 May 2026 that it is testing new formats there, and on 14 September 2026 an observer spotted sponsored text links under a Sponsored label. Google has not announced that text-link test.

Is there a way to turn off ai ads?

We could not verify one. Google Ads Help says advertisers cannot opt out of serving ads in AI Overviews, and no setting is documented for the AI Mode text-link test. For ChatGPT, OpenAI's ad policies describe placement rules, not a user or advertiser off switch.

Do ai-generated ads actually work?

There is no published performance evidence for these units that we could find. Barry Schwartz doubts the text-link format will match the click-through of normal ads, but that is opinion. Test with a baseline and a holdout, and attribute any result to your own data.

Is ai advertising illegal?

Not by itself. Existing ad rules still apply: the FTC says an ad is deceptive if it is not readily identifiable as an ad, and finance, gambling and health carry sector rules. Ask counsel about your market; this is not legal advice.

How can I tell whether ads in AI Mode Google is testing have touched my account?

You probably cannot yet. Google Ads Help says there is no segmented reporting for ads in AI Overviews, and nothing published covers the AI Mode text-link test. Keep a weekly baseline, annotate changes and compare search terms before and after.

What are conversational discovery ads?

Conversational discovery ads are one of two formats Google said on 20 May 2026 it is testing in AI Mode. The ad answers a person's specific question with creative built by Gemini, alongside an independent explainer, and stays labelled Sponsored.

Can regulated brands run google ads in ai mode?

Google Ads Help says finance, gambling and healthcare ads are not shown in AI Overviews. We found no equivalent statement for AI Mode. Assume the strictest reading, keep landing pages self-contained and confirm with Google before spending.

Keep reading

Read more on this topic

Ready to outfox the label problem before it reaches your account?

folkfox audits regulated paid search accounts for claim-carrying copy, self-contained landing pages and measurement baselines, so a test of google ai mode ads never becomes a surprise.

Want folkfox in your Google results and AI answers? Set folkfox as a preferred source.

The den

Where to?

Pricing

Choose a section. Enter opens it, Escape continues reading.

Cookie preferences

folkfox uses data the way we use strategy: only when it earns its place.