BEVIZIO

AI visibility for accounting firms: does ChatGPT name your firm?

Dated sample: October 7, 2026

Someone can ask ChatGPT, Gemini or Google AI Mode a question such as “Who are the best tax preparers in Houston, Texas?”, and the answer can list businesses. Bevizio asks such questions for CPA, tax and bookkeeping firms, again and again, and counts how many answers name yours.

The free check asks for no account. Pricing starts at $29 a month for Growth, with Pro at $79, and both begin with a 14-day trial that takes a card; cancel before it ends and you pay nothing. See the plans side by side on pricing.

Ask the same question twice and the list can change, so the dependable way to see how your firm fares is to ask many times and count. In the sample below, every one of ChatGPT’s 78 answers listed at least three firms. Bevizio is a measuring tool. It cannot change what an assistant says, and it makes no promise about results. Trying to influence those answers is often called GEO; the background is in generative engine optimization explained.

What you would see each week

You enter your firm, the services you sell and the towns you serve. Every service-and-town pair becomes a question worded as a client would type it, such as bookkeepers in Chicago. A question is put to the assistants 8 times a week: three times to ChatGPT, twice to Gemini, twice to Google AI Mode and once to Claude.

  • Is your firm named? A running tally per question and per assistant, such as “named in 2 of 3 answers”. While the numbers are small, you see the raw counts and never a percentage.
  • Who else is shown? The other businesses ChatGPT listed in its answers, alongside the stars and review totals it displayed for them. Gemini and Google AI Mode answers carry no business cards at the moment, so for those you see the sources instead.
  • Which sites does the answer lean on? The pages behind each answer, sorted into maps listings, business directories, review sites and everything else. Directories and maps are treated as sources and are never counted as competitors.
  • The picture by service and by town. A firm that sells both tax work and bookkeeping, or that runs two offices, can see which pairings get named and which do not.

Everything shown is an observation: counts of real answers, with no advice and no score. The example below uses an invented firm and made-up numbers.

Check your own firm

The free check sends a single question to ChatGPT, Gemini and Google AI Mode, built from one of your services and one of your towns. There is no sign-up, and your firm’s name is left out of the question. You get a dated sample of who the answers named, which is a snapshot and not a verdict on your firm. The trial follows your firm every week.

Growth tracks 20 questions for $29 a month. Pro tracks 75 for $79, although a Pro trial follows only 20 until the first payment. On either plan each question gets the same 8 weekly asks. The trial needs a card and rolls into a paid plan unless you cancel in Settings. Cancel before the trial ends and you pay nothing; if you forget, email us within 7 days of the first charge for a full refund. After that you can cancel a paid plan at any time, and it runs to the end of the period you paid for. Paddle sends a reminder at least 7 days ahead of the first charge, and the refunds page has the full terms.

Bevizio measures dated observations. It does not rate, rank, refer or advise.

Who this is for

Partners, owners and office managers of accounting, tax and bookkeeping practices who want a weekly record of how the assistants treat their firm. A one-person tax practice and a firm with several offices both fit, because each extra service or town is just another question. An account holds one firm name, so a network of separately branded firms is out of scope for now. The advertising rules quoted below govern CPAs; a bookkeeper or a preparer who is not a CPA falls under other rules that we have not read.

What we saw on October 7, 2026: a dated sample

We gave ChatGPT, Gemini and Google AI Mode a set of US accounting questions on a single day and counted the responses. The sample tallies the firms and sources in the answers; it does not test whether a specific firm was named. Three things stood out. ChatGPT listed firms in every answer. Directories supplied a large part of what ChatGPT and Google AI Mode cited: 36% of ChatGPT’s outside citations and half of Google AI Mode’s. And most of Gemini’s outside links were Google Maps listings. One day of observations cannot rank any firm, and it is not advice on getting named.

How we ran it

  • Date: October 7, 2026.
  • Assistants: ChatGPT, Gemini and Google AI Mode, collected through a data provider, by the same route Bevizio uses for its own checks. The country was set to the United States, and each city’s Google location was passed on wherever the assistant could use it.
  • Questions: Five services across eight cities gave 40 questions. Every assistant received every question twice, so 240 asks went out. Two failed and are excluded, which leaves 238 answers.
  • Services: tax preparers, bookkeepers, small business accountants, CPAs and accountants for self-employed taxes.
  • Cities: large ones were New York, Los Angeles, Houston, Chicago and Miami; Columbus was mid-size; Boise and Omaha were smaller. We decided the services, the cities and the source lists before running anything and did not adjust them afterwards; the one group added later is labelled below.

The sample rates no firm and this page names none. Firm names do appear in the raw answers, so those stay private and we publish only counts. How the collection works is explained on how we measure.

What the answers listed and cited

An “outside citation” is a link to a site other than the assistant’s own pages, and a single answer can contain several. We counted a site as a directory when its web address matched a list of business directories, review and ratings sites and lead marketplaces drawn up before the run. Google Maps links are counted separately and are not treated as directories. Everything else, including firm websites and publishers, is filed under “other” and left unsplit.

AssistantAnswersOutside citationsDirectory citationsAnswers citing a directoryGoogle Maps links
ChatGPT7822682 (36%)62 of 780
Gemini8031825 (8%)11 of 80180 (57%)
Google AI Mode8019497 (50%)45 of 800
  • Most of ChatGPT’s directory citations, 65 of the 82, came from a single site.
  • Neither ChatGPT nor Google AI Mode produced a Google Maps link here, so those zeros are genuine.
  • Of ChatGPT’s 226 outside citations, 5 (2%) pointed to tax authorities or other government sites.
  • Google AI Mode also produced 466 links that lead into Google’s own viewer. They are not outside sources, so the 194 excludes them.

Chosen after the run, so a second look and not part of the protocol: scanning the sites the answers cited, we spotted sites that list, rank or review accounting firms and that our lists had missed. We grouped seven of them after the fact. They turned up in 38 of ChatGPT’s 78 answers (40 of its 226 outside citations, 18%) and in 11 of Gemini’s 80 answers (48 of its 318, 15%), and barely in Google AI Mode (3 of its 194). Because we formed the group after seeing the data, it points at something worth measuring and does not count as a finding. It also means the fixed-list directory shares in the table are a floor for how much the answers lean on sites that list firms.

What ChatGPT listed

ChatGPT was the only assistant whose firm cards, the business entries it places inside an answer, we could count. We did not read Gemini or Google AI Mode answers for firm names, so nothing is said about how many firms they listed.

  • Firm cards appeared in all 78 of its answers, and no answer had fewer than 3.
  • The median answer carried 7 cards, with a range of 3 to 12.
  • Two answers to the same question mostly showed the same cards (median overlap 0.82 across 38 questions; 2 questions drop out because one of their answers failed). We measure overlap as the share of the shorter answer’s cards found in the other, so 1 would mean a complete match. That is one day, and it tells you nothing about next week.

How far to trust this

On these questions ChatGPT named specific firms every time and cited a directory in most of its answers, while Gemini mostly pointed to Google Maps listings. We never checked whether any firm was listed on those sites or whether a listing makes an assistant name it, and the sample does not say it does.

  • It covers one day and three assistants, with 5 services in 8 cities, five of them large. Other questions and smaller towns could look different.
  • Each question was asked twice, so these are snapshots and not rates.
  • We infer a site’s type from its web address and a fixed list, which means some sites will be filed wrongly.
  • A data provider supplies the answers, and they can differ from what a person sees in the ChatGPT, Gemini or Google app itself.

The questions Bevizio asks for an accounting firm

Here is the kind of question Bevizio sends, and the five used in this sample, worded as a client might type them. The wording is the client’s and says nothing about any accountant. What we measure is the assistant’s reply.

  • “Who are the best tax preparers in Houston, Texas?”
  • “Who are the best bookkeepers in Chicago, Illinois?”
  • “Who are the best small business accountants in Columbus, Ohio?”
  • “Who are the best CPAs in Los Angeles, California?”
  • “Who are the best accountants for self-employed taxes in Boise, Idaho?”

In your account a question is a service plus a town. For the free check, your firm’s name stays out of the question. Bevizio then tallies how often the reply includes your firm, what other businesses it lists and which sites it points to.

Questions accountants ask

Is this only for CPA firms?

No. Any service in any town can be a question, so a tax preparer, a bookkeeper or a full-service CPA firm can use it. Only the rules section further down is specific to CPAs.

Does Bevizio handle client financial data?

No. It does not ask for tax returns, books or any other client information. It puts public questions to the assistants and stores the replies, plus the firm details you enter.

Can one account cover several offices?

If the offices share one firm name, yes: each service and town is a question, up to 20 on Growth and 75 on Pro. Firms with different brand names need an account each, and a combined account for a group is not available yet.

What does 20 questions get me?

A question pairs one service with one town, so 20 could be 4 services in 5 towns. Growth suits a firm with a few services and offices; Pro goes up to 75.

How should I read “named in 2 of 3 answers”?

That week, two of the three answers to that question from that assistant included your firm. It is a tally and not a score. Answers shift, so Bevizio keeps the plain counts while they are small and follows the trend over weeks.

Does Bevizio list firms or pass on leads?

No. There is no public list of firms, no referral, and no fee for sending anyone anywhere. In your account and in the free check you see the other businesses ChatGPT listed, as it gave them; for the other assistants the free check shows other names found in their sources.

May we put these numbers in our marketing?

Bevizio does not advise on that, and it offers no badge or shareable “AI named us” graphic. Advertising rules for accountants apply to the firm and vary by state, so read the rules note below and ask your state board or counsel.

Does watching the numbers make us show up more?

Not by itself. Bevizio records what the assistants say and has no way to change it. What is known about what moves them is set out in why AI engines name your competitor instead of you.

What the rules say

Advertising rules for accountants are aimed at the accountant, and states differ. We read three primary texts on October 7, 2026: the AICPA Code of Professional Conduct, California and New York. This page names no firm, says nothing about any firm being named, and is not legal advice. The AICPA Code binds AICPA members in public practice, and the two state texts bind licensees. New York extends its section 29.1 to public accountancy through section 29.10.

“A member in public practice shall not seek to obtain clients by advertising or other forms of solicitation in a manner that is false, misleading, or deceptive.”

AICPA Code of Professional Conduct, ET 1.600.001, paragraph .01, read on October 7, 2026. Source 1 below.

“Promotional efforts would be considered false, misleading, or deceptive if they … create false or unjustified expectations of favorable results.”

AICPA Code, ET 1.600.010, paragraph .02(a), read on October 7, 2026. Source 1 below.

“A licensee shall not advertise or use other forms of solicitation in any manner which is false, fraudulent, misleading, or in violation of Section 17500 of the Business and Professions Code.”

California Code of Regulations, title 16, section 63, read on October 7, 2026. Source 2 below.

“[A]dvertising or soliciting that … makes claims of professional superiority which cannot be substantiated by the licensee, who shall have the burden of proof”

New York 8 NYCRR section 29.1(b)(12), advertising or soliciting not in the public interest, read on October 7, 2026. Source 3 below.

Other states have rules of their own that this page does not cover. Bevizio does not interpret any of these rules for a firm, and nothing above tells a firm what it may publish.

Scope

Bevizio measures dated observations. It does not rate, rank, refer or advise. It is not an accounting firm, a marketing consultancy or a client referral service, and this page gives no advice on advertising an accounting practice.

Next step

Start with the free local check, or begin the trial to follow your firm weekly. For the wider picture, see the local business overview, how we measure and what AI visibility is and how it is measured. Similar dated samples cover US law firms and US dental practices. To try the check by hand first, read how to check whether ChatGPT names your local business.