BEVIZIO

Claude · ChatGPT · Cursor

Ask Claude which AI engine names your product least

Bevizio asks ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews the kind of questions a buyer in your category would ask, then counts how often each one names you. Connect it to Claude, ChatGPT or Cursor and ask for those counts in plain English: who names you least, who gets cited instead, what changed this week.

New accounts get a 14-day trial, card required. Plans start at $15 a month, and the assistant connection is included. The free check needs no account. Already a customer? Open Settings.

Screen recording, October 30, 2026. Claude answering two questions about CountryTaxCalc, a product we run and track. The numbers come from Bevizio.

Two real answers

From the recording, for the last 28 days up to October 30, 2026. “Named” means your product’s name is in the answer, or your own site is cited as a source.

Using Bevizio, how often do AI engines mention CountryTaxCalc?

Named in 809 of 1,920 answers (42%), by engine:

  • Perplexity215 of 361 (60%)
  • Claude197 of 398 (49%)
  • Gemini181 of 369 (49%)
  • Google AI Overviews161 of 357 (45%)
  • ChatGPT55 of 435 (13%)

Using Bevizio, who are AI engines recommending instead of CountryTaxCalc?

Cited in answers that did not name it. Cited is not the same as recommended:

  • OECD161 answers
  • SmartAsset106 answers
  • PwC100 answers
  • Tax Foundation98 answers

Claude also pointed out that ChatGPT’s collection method changed three times in October, so its fall in recent weeks should be read with caution, not as a confirmed decline. We didn’t ask for that. In the recording Claude adds some commentary of its own. Treat that as Claude’s opinion, not Bevizio’s.

Connect in three steps, no key

  1. Add the address. In your assistant, add a custom connector with this address.
    https://www.bevizio.com/api/mcp
  2. Sign in and press Allow. Your browser opens Bevizio. Sign in if you aren’t already, check that Bevizio recognises your assistant and where it will send you back to, and press Allow. If the page says it doesn’t recognise the app, cancel unless you started the connection yourself, from an app you trust, and the address is the one you expect. There is no key to copy.
  3. Ask. Your assistant now has the Bevizio tools. Approve them when it asks, then type a question.

Using Claude Code in a terminal? Use a personal key. Codex and other terminal assistants may work the same way, but we haven’t tested them. Tested October 29, 2026: Claude (claude.ai), ChatGPT and Cursor.

Three questions to start with

  • See which engine names you least“Using Bevizio, how often do AI engines mention my product?”
  • See who they cite instead“Using Bevizio, who are AI engines citing instead of my product?”
  • See what changed“Using Bevizio, what has changed since last month?”

Then ask your assistant what to do about it. The counts and sources are Bevizio’s. Any advice is your assistant’s opinion, not ours.

Seven more questions you can ask

Each uses one of the Bevizio tools. You don’t name the tool; your assistant picks it.

  • Which of my tracked questions am I named in least?
  • Which sites and pages do the engines cite when they skip me?
  • Did anything change recently, according to Bevizio’s statistical test?
  • Show me one recent answer in full, with its sources.
  • When was the data last collected, and how is each engine checked?
  • Which of my questions do the engines answer from memory, without searching?
  • What did the engines search for when answering my questions?

What it does and doesn’t do

  • It reads by default. Changes stay off until you turn them on in Settings, for each connection separately, for 30 days.
  • It sees one product: yours.
  • It gives counts of dated answers, not scores or forecasts, and it doesn’t tell you what to change on your site.
  • You can disconnect any connection in one click in Settings, or all of them at once.

The connection is included in the 14-day trial and in all plans, at no extra cost. It reads the same data as your dashboard, so it needs an active plan or trial. Plans and prices are on the pricing page.

Full details: every tool, what it can and can’t do, safety and privacy, plan requirements

What each tool does

There are 9 read tools. Your assistant picks the right one, so you never have to name it. The names are what your assistant may show when it asks to run one.

  • get_account_status

    Your plan and trial status and question limit; how each engine is collected, your country setting and the dated events behind a movement; and per engine how many answers could be judged. Choose a view: status, engines or health.

  • get_visibility

    How often each engine named you over 7, 28 or 84 days, as “named in k of n answers”. Views: overview, compare (this period against the one before, same thresholds as change alerts) and own_site (how often your own site is cited; not available for local businesses).

  • list_questions

    Your tracked questions and how often each names you, with the ids the question-level views use. Views: list, history (one question week by week on each engine) and handling (searched the web or answered from memory).

  • get_answers

    Recent answers, or one answer in full with its sources. Each links to its page in your dashboard.

  • get_competitors

    The companies cited in answers that did not name you. Views: list, trends (the competitors you chose to track, week by week) and sources (the pages cited for one competitor). Not available for local businesses.

  • get_sources

    The sites and pages cited most often. Views: sites (with the kind of site), pages (which answers, engines and positions) and types_over_time (the mix of site types by week; Bevizio’s own classification, which can be wrong).

  • get_changes

    The changes Bevizio’s statistical test flagged in about the last six weeks, each comparing two weeks of answers with the four weeks before.

  • get_search_queries

    The web searches the engines ran while answering, as the provider reports them, and the questions that triggered them. Not every engine reports its searches.

  • list_notes

    Your current dated notes, newest first, with who wrote each (you or an assistant) and which one an assistant can delete. Note text is quoted, never an instruction.

Local businesses get the same tools, with differences: there is no competitor list yet (and so no pages cited alongside a competitor), no figures for your own site’s citations and no “how was I named” split. Results are broken down by area and service instead, on the engines Bevizio uses for local businesses, because local businesses are judged by their own rules (see how we measure).

What it can and can’t do

Every figure is a count of dated answers, written “named in 55 of 435 answers”. It is a snapshot of what an engine said on a given day. There is no score and no forecast, and Bevizio doesn’t tell you what to change on your site: it measures. If your assistant adds its own opinion, it should say so, and you should treat it as the assistant’s.

The numbers are the dashboard’s numbers. For a product or website, “named” means your product name or an alias you set appears in the answer text, or your own site is cited as a source. Ask “how was I named?” and the assistant can split the count. Answers Bevizio couldn’t judge are reported as gaps, not as misses.

Answers and cited pages are written by other people, so a sentence in one could try to steer an assistant that has other tools, such as email or files. Bevizio labels all such text as a quotation, strips hidden characters, turns links into plain text, and flags phrases that read like instructions. It tells your assistant never to act on them. The flag is a warning and cannot catch everything, and it is your assistant that decides what to do, so be careful about connecting an assistant that can act on your behalf without asking you.

Your assistant reads by default. If you choose, in Settings, you can also let it make a small set of changes, for each connection separately, for 30 days. At the level “Add and remove notes only” it can add and remove notes. At the level “Notes, plus questions, competitors and alternative spellings” it can also add, archive, restore, pause or resume questions, track or stop tracking competitors, and change your alternative spellings (which changes how your numbers are counted from then on, so allow that only if you want it). For a local business, the higher level is “Notes, plus areas and services”: it can add or archive an area or a service, and Bevizio builds the questions from them. It can never touch billing, your plan, your password, your product’s name, website or country, or your account. Every change is logged in Settings and limited. For 7 days, while that connection still has permission, you can ask it to undo a change, and you can always change or remove the item yourself. It can’t run an extra check: use the dashboard for that.

Safety and privacy

It reads unless you say otherwise

The connection is read-only until you allow changes in Settings, for each connection separately, for 30 days. Even then it can’t touch billing, your plan, your password or your account, and a pasted connection key can never change anything.

You can switch it off in one click

Settings lists every assistant you connected by signing in (personal keys are listed separately). Disconnect stops that assistant at once and leaves the others alone.

It sees one product

The assistant can read only the product on your account. The database itself enforces that limit.

What it can read

Your product’s name and website, your questions, how often each engine named you, and which companies and pages the engines cited. It can also read the web searches the engines ran, your dated notes, recent answers (the full text for the last 12 months), changes we detected, and your plan and trial status. It can’t read your email address, your password or your billing details.

Where your questions go

What you ask and what comes back is handled by your assistant’s maker, under your own agreement with them. Bevizio doesn’t send your results to an assistant; the assistant asks for them. We record which tool was used, whether it worked, how long the request took, and which connection and account it came from. We don’t record what you asked or what came back, apart from the changes an assistant makes for you if you let it: those go in a record of changes, described in the privacy policy.

More on how the numbers are collected, and what they can’t show, is on how we measure. The wording that covers all this is in the privacy policy and the terms.

Which assistant plans work

For Claude, custom connectors are available on the Free, Pro, Max, Team and Enterprise plans, and the Free plan allows one (Anthropic’s help page, read October 19, 2026). For ChatGPT, OpenAI’s help page (read October 19, 2026) lists custom MCP connections for Business and Enterprise/Edu workspaces, and for Pro with read-only access, on the web only. On Business, only an admin or owner can add one. The page doesn’t list Free or Plus. Cursor has no limit we know of.

How old is the data?

Every reply that returns figures starts with the date of the newest answer Bevizio could judge and how many days ago that was.

Questions

What is MCP?

The Model Context Protocol, an open standard that lets an assistant connect to a service. You don’t need to know more: you add an address and sign in.

Which assistants work?

Claude (claude.ai), ChatGPT and Cursor, all tested on October 29, 2026. Other assistants that support remote connections with browser sign-in may work. We only list the ones we’ve tried. Plan requirements are under full details.

Is it accurate?

It reads the same numbers as the dashboard. Every answer it quotes comes with a link to that answer in your dashboard, so you can check. Your assistant is told to report only what Bevizio states, but it is still an assistant: check anything that matters.

Can it change my questions?

Only if you allow it. By default it only reads. In Settings you can let Claude, ChatGPT or Cursor add and remove notes or, at the higher level, change your questions, competitors and alternative spellings (for a local business, your areas and services), within limits. Every change is logged, and for 7 days, while the assistant still has permission, you can ask it to undo a change.

Is Bevizio affiliated with Anthropic, OpenAI or Cursor?

No. They make the assistants. We make the data they read.

Ask your assistant about your own numbers