Tobias Lochau, Seofable·2026-10-29·7 min read

Google AI Mode Named Only 11% of 1,035 Agencies for Their Own City (Study, October 2026)

Every agency we talked to this autumn asked the same question in different words: "Does Google's AI even know we exist?" Nobody had a number. So we measured it.

What we did

We took 1,035 marketing, SEO and web design agencies from two lists we had already built for outreach: 383 agencies in 12 UK cities and 652 agencies in 15 US cities. For each agency we kept three fields only: domain, agency name, city.

For every city we asked Google AI Mode four questions, phrased the way a business owner would type them:

• best marketing agency in {city}
• best seo agency in {city}
• best web design agency in {city}
• top digital marketing agencies {city}

That is 108 answers, pulled on 28 and 29 October 2026 through DataForSEO's AI Mode endpoint, desktop, with the location set to the UK or the US. We then checked every answer for each agency's domain and, separately, for its full name. A name only counted when it was distinctive. "Bristol Marketing Company" appearing in an answer about Bristol marketing companies proves nothing, so names built only from generic words were counted by domain alone. Matching used word boundaries, so "Fifteen" did not match "fifteen years of experience".

The headline number

117 of 1,035 agencies, 11.3%, are named by AI Mode for at least one of the four questions about their own city.

Read that the other way round: nine out of ten agencies on a list of real, active, outreach-worthy agencies do not appear when a buyer asks Google's AI who the best agency in town is.

A looser count that accepted any substring match gave 124 agencies (12.0%). We publish the strict number. Matching ignores trademark symbols such as ® in agency names; one Austin agency was found only after that fix, which is why the first published draft said 116.

The number swings a lot by city

Nottingham is the outlier. 9 of 34 agencies there get named, 26%. Leeds, with a sample of the same size and a bigger market, sits at 5%. We did not find a pattern in the data that explains this. Our best guess, and it is a guess, is that the directories AI Mode leans on have fuller, better-reviewed Nottingham listings. If you run an agency in Leeds, that is an open goal.

What AI Mode actually cites

This is the part that changes what you should do on Monday.

AI Mode is not reading 40 agency websites and forming an opinion. For half of these questions it reads a directory, mostly Clutch, and summarises what it finds there. Only a third of the answers link to any agency's own site, and when they do it is usually one site, not a list.

The agencies that got named by domain, 18 of the 117, are the ones whose own website was good enough to be the source. The other 99 were named because a directory named them first.

What this means if you run an agency

1. Your Clutch profile is now a ranking factor for AI answers. Not metaphorically. In 42 of 108 answers it is the source. Complete it, get reviews on it, keep the service list and the city accurate. Thumbtack matters in the US, Digital Agency Network and DesignRush in the UK.
2. Make your own site the kind of source AI can quote. The 18 agencies named by domain all have one thing in common: the answer could lift a sentence from their site that says what they do and where. Service plus city in plain text on the page, not in an image or a slogan.
3. Use the same name everywhere. Our matching found agencies that are listed under one name on Clutch and another on their own site. A model that sees two names has two weak entities instead of one strong one.
4. Check your own city before you spend money. The four questions above cost nothing to type into Google. If you are not there, you now know why, and it is fixable without a new website.

What did not show up

Nothing in these 108 answers linked to an agency blog post, a case study page or a LinkedIn profile as the reason an agency was named. If your plan for "AI visibility" is publishing more articles on your own domain, this data does not support it. For this type of question, directory presence beats content volume.

Limits of this study

• One measurement on two days. AI Mode answers change; we will repeat it.
• Four questions per city. Different phrasing ("affordable", "small business", "near me") may produce different lists.
• Desktop only, location set through DataForSEO, no personalisation. Your own result will differ from ours.
• Name matching misses agencies that rebranded or trade under a different name. The true share is probably a little higher than 11.3%, not lower, but we have no reason to think it is double.
• The sample is agencies we had already qualified for outreach, so it skews towards agencies that have a website and a reachable email. Agencies without a website would do worse.

Update, 1 November 2026: the second measurement

We ran the same 108 questions again on 1 November, three to four days after the first wave. 104 of the 108 city and question pairs returned a usable answer in both waves (three Nov requests failed, one Oct), so the comparison below uses only those 104 pairs.

Only 69 of the 116 agencies named in the first wave, 59%, were named again a few days later. We do not know yet whether that is day-to-day variation in AI Mode or real change in its sources. Two measurements cannot separate the two. What they can say: a single check of "are we named in AI Mode" is a snapshot, not a status. We will repeat the measurement on the first of each month with the same cities and questions and publish the table here.

The data

• Per-city table: google-ai-mode-agencies-2026-10-cities.csv
• All 108 answers with the sources they link: google-ai-mode-agencies-2026-10-answers.csv

We did not publish the agency list itself. If you want to know whether your agency was in the sample and whether it was named, email us the domain and we will tell you.

FAQ

Is this the same as Google AI Overviews?

No. AI Mode is the conversational search tab Google rolled out in 2025; it answers with a longer text and a small set of links. AI Overviews are the summary box above classic results. We measured AI Mode only.

Why did you use DataForSEO instead of typing the questions into Google?

108 questions across two countries with a fixed location are not reproducible by hand. The endpoint returns the same answer text and links a user would see, with the location pinned. The cost was 43 cents.

Does being named by AI Mode bring clients?

We cannot tell from this study. It measures presence, not clicks. What we can say is that an agency that is not named cannot be chosen.

Can I get the list of named agencies?

We do not publish other companies' names from an outreach list. You can run the four questions for your city yourself in a few minutes.

Will you repeat this?

Yes, monthly, with the same 27 cities and the same four questions, so the numbers become comparable over time.

✓ Fact-checked 2026-10-29, Korrektur 20:15 (Namensabgleich ignoriert ®, 116 → 117); Update 2026-11-02 mit zweiter Messwelle vom 01.11. (104 gemeinsame gültige Paare, 69 von 116 erneut genannt), Rohdaten data/study/ai_mode_raw_2026-11-01.json. All numbers come from our own dataset of 108 Google AI Mode answers pulled through DataForSEO's AI Mode endpoint on 28 and 29 October 2026; the raw answers, the per-city table and the matching script are published below. External facts checked: Clutch, Thumbtack, DesignRush, The Manifest, Sortlist and Digital Agency Network are B2B directories with agency listings and reviews; Semrush Agency Partners is Semrush's own agency directory. No market-size or revenue claims.