Free AI Visibility Checker: Does AI Recommend Your Brand?

Current tool, 28 October 2026, 14:43 UTC. Category found: “all-in-one SEO and GEO platform”. Claude answered all 6 questions and Gemini 1 (Google quota), so 7 answers were scored. Score 2/100: named in 0 of 7. These are the 6 questions the checker wrote and the brands the first answer to each listed, in order.

Two lessons from our own result. First, a model answering from training data, with no web search, names the incumbents of the wider category: no dedicated AI-visibility tracker appeared, even for questions about AI citations. Second, the score is honest about a young brand. Our earlier check on 1 Oct 2026, answered by ChatGPT alone, gave the same 2/100. We publish ours because the method should hold for everyone, us included.

Current tool, 28 October 2026. The checker uses one model to find your category and write the six questions, then sends those questions to the answer models. The two can differ; today Claude Haiku 5.5 does both. The steps, in order, with the same settings for every brand:

The scoring formula

• Each answer scores 90 if you are named 1st, 75 2nd, 60 3rd, 48 4th, 35 5th or lower, and 30 if you are named outside a numbered list; +6 if the answer cites your site. Not named: 3.
• We average the answer scores and round that average to a whole number. Your score is 70% of that average plus 30% of the share of answers that name you, rounded to a whole number. This is how the checker’s code calculates it.
• The +6 citation bonus is added to an answer’s score before averaging, and needs an assistant with web search. The assistants answering today don’t search the web, so today the best answer scores 90 and the maximum score is 93 (0.7 × 90 + 30).
• Only answers that came back count. Claude answers all 6 questions, so at least 6 answers are scored; each Gemini answer that gets through adds one, up to 12. The report says how many answers each assistant gave.
• Example: named 2nd in 2 of 6 answers and absent from 4 gives (2 × 75 + 4 × 3) ÷ 6 = 27 average, so 0.7 × 27 + 0.3 × 33.3 = 28.9, rounded to 29.

Which models answer

Paused assistants are skipped and left out of your report. Historical sample, 19 Jun – 20 Oct 2026 (146 visitor checks, our own tests excluded): ChatGPT answered all 146; Claude also answered 19 of them; Perplexity was tried once and returned nothing (Rankfender’s own data). Rankfender’s paid plans use a separate, larger set of assistants.

You can repeat what the checker does in 20 minutes, with the apps your buyers use. It also shows you what the API cannot: answers with web search on.

This is the same method as the free checker, with web search on. Use both: the checker for a quick, repeatable baseline; the manual check to see the sources assistants cite.

The bands below come from the formula above. The share next to each band is how the 122 brands checked so far are spread (Rankfender’s own data, latest check per brand).

In this sample, no brand landed in 15–39. That reflects these brands, questions, models and scoring rules, not a universal distribution of AI visibility: most of these brands were absent from every answer (score 2–3) or named in most of them. A brand named 2nd in 2 of 6 answers would score 29.

Two factors influence AI visibility. Assistants answering without web search rely heavily on what they learned in training, along with the model’s current behaviour. Search-enabled assistants can also find and cite relevant pages from the web. The moves below can help; the impact varies by assistant, and they are not guaranteed to change your score immediately.

Competitor experiment, 28 October 2026. Free checkers measure three different things: some ask an AI live, some look your brand up in a stored database of past answers, and some scan your site without asking any AI. We opened 13 of them on 28 October 2026 from a US location, without an account or email, and recorded what each one really does.

Tested on 28 October 2026; “stated” means as written on the vendor’s page. Backlinko’s checker (Semrush) and SEO Review Tools were read but not run (captcha). SE Ranking has no free checker; its ads lead to a 10-day trial of SE Visible.

Show the dated screenshots (4)+

When one snapshot is not enough, a tracker asks your own prompts on a schedule. Here are the 12 trackers that free-checker pages push, from the cheapest entry plan up, checked on each vendor’s own pricing page on 28 October 2026. Prompts are the number of questions tracked; the same prompt is usually asked to every included assistant.

Prices in US dollars as shown on each vendor’s US pricing page on 28 October 2026, before tax. “Yearly” is the monthly equivalent when billed yearly. Semrush One Starter ($199, 50 prompts) and Amplitude (AI visibility on its free plan, ChatGPT and AI Overviews) were also checked. Peec’s yearly price and Scrunch’s frequency are not published.

Show the dated screenshots (2)+

Historical sample. 146 free checks run by visitors between 19 Jun and 20 Oct 2026, our own 16 tests excluded. ChatGPT answered all 146 (876 answers); Claude also answered 19 of them (114 answers): not today’s configuration. For brands checked more than once, only the latest check counts (122 brands). Aggregates only: no brand is named.

Key findings

• 89%

  Most brands are not named at all

  108 of 122 brands checked with this tool were named in none of their 6 answers (Rankfender’s own data, latest check per brand, 19 Jun – 20 Oct 2026). Median score: 2 out of 100. A low score is the normal starting point, not a fault in your site.

• 2/100

  Including us

  Rankfender ran its own domain through the current checker on 28 October 2026: named in 0 of 7 answers (6 from Claude, 1 from Gemini). Semrush and Ahrefs were named in all 7, Moz Pro in 5. See the questions and answers.

• 6 questions

  Six questions, about 30 seconds

  The current checker asks 6 buyer questions to Claude through its API: 6 answers per check, up to 12 when Gemini’s quota allows. The formula, the models and what is not measured are all on this page.

• API ≠ app

  What a free check cannot see

  API answers have no web search, no memory and no location, so they can differ from what someone sees in the ChatGPT or Gemini app. Limits explains what that means for your score.

• 95 · 30 · 0/5

  Free checkers disagree

  The same site, rankfender.com, on the same day: 95/100 on a site-setup scanner, 30/100 on HubSpot’s grader, named in 0 of 5 GoBIG searches. Most free checkers don’t ask an AI live at all. 13 checkers compared.

• Never named (latest check per brand)89% (108 of 122)

  Named in none of their 6 ChatGPT answers. Median score 2 out of 100.

• Named in every answer (latest check per brand)7% (9 of 122)

  6 of them scored 70–100; 3 scored 40–69, named every time but lower in the lists.

• ChatGPT answers that named the brand (all 146 checks)9% (81 of 876)

  gpt-4o-mini, 146 checks × 6 questions.

• Claude answers that named the brand (19 checks)11% (13 of 114)

  Claude Haiku 4.5, the 19 checks it also answered, × 6. Small sample.

• Checks where an answer cited a source (all 146)0% (0 of 146)

  API answers without web search cite nothing; only Perplexity returns sources, and it never answered.

Read it as a baseline, not a benchmark: visitors who run a free checker skew towards small and new brands. The pattern still matters. Assistants answering from memory repeat a short list of well-known names, so a new brand is usually absent rather than ranked low.

Use the score to see whether you are in the answers at all and who is in your place. Do not use it as a precise measure. Here is why.

API answers are not the app

The checker calls each model’s API without web search, memory or location. In the app, OpenAI says ChatGPT may search the web, rewrites the question and uses saved memories and an approximate location, so a person may see a different answer. How much depends on the app and the features switched on.

Answers vary between runs

Ask the same question twice and the list can change. SparkToro’s 2,961-run study found under a 1-in-100 chance that two runs return the same list of brands. With 6 to 12 answers, one check can’t tell a 25% mention rate from a 50% one. Repeat checks over weeks.

Six questions, written by a model

The questions are generated for your category. They may not match the exact words your buyers use; check the list in your report.

Brand-name matching

We look for your brand name as written. Abbreviations, product names or misspellings in an answer are not counted.

No sources yet

Without web search the answers cite nothing, so the free check cannot show which pages assistants rely on.

A snapshot, not tracking

For weekly or daily answers from more assistants, with your own prompts, you need a tracker. Paid trackers compared.

Disclosure: We make Rankfender and this checker. The other free checkers were tested on their public pages on 28 Oct 2026 without paid accounts, and paid-tracker prices come from each vendor’s own pricing page that day. No affiliate links.

• Our tool: the checker’s source code was read on 28 October 2026 (category step, prompts, scoring, models). The formula on this page is that code.
• Claims gate: a live check was run on 28 October 2026; only assistants that answered it are named.
• Our data: Rankfender’s own data, from the checks stored by this tool between 19 Jun and 20 Oct 2026: aggregates only, our own tests excluded, latest check per brand.
• Other checkers: each free tool was opened on 28 Oct 2026 from a US location, with dated screenshots. We did not enter an email or create an account; where a tool needs one, the table says so.
• Paid trackers: prices, prompts, assistants and update frequency from each vendor’s own pricing page on 28 Oct 2026, in US dollars.

Evidence

What each claim on this page rests on.