6 AI platforms. Your prompts. Every answer on the record.
AI Visibility Audit for AI Search
An AI visibility audit asks ChatGPT, Google AI Overviews, Google AI Mode, Gemini, Perplexity and Claude the questions your buyers ask, then scores what comes back from 0 to 100. You see every answer, the exact prompt we sent, and the sites each engine cited.
This is the audit at the core of 99Visibility, an AI search visibility tool built to find what is wrong and say how to fix it. Each prompt is asked on each platform in your plan, and every number on the report is counted over those answers.
We send your prompt exactly as you typed it: no brand name, no instructions, no country added. That is what a buyer types, so it is what we measure. Google AI Overviews and Google AI Mode have no API, so we read what Google shows for each in your market.
A mention is not a recommendation. Each answer records whether you were named, the tone around your name and its links; with competitors tracked, it also records where you sat in the list.
The same run checks whether AI crawlers can reach your site, so the fix list covers access as well as answers.
37/100
+2 since last auditNamed on 3 of 6 platforms× 0.70 coverage
Visibility Over Time
Overall visibility score across audits
What the score measures
Mention rate (40%)
Mention rate is the share of answers that name your brand. Each prompt asked on each platform is one answer: 20 prompts on 6 platforms is 120 answers, and if 18 of them name you, your mention rate is 15%. Your brand is matched with its spacing variants and your domain name, so "99 Visibility" and "99visibility" both count.
Accuracy (30%)
Accuracy starts at 100 and drops with each statement AI gets wrong against the facts in your brand profile, weighted by how serious the worst error is and how often errors occur. With no facts on file there is nothing to check, and it stays at 100.
Sentiment (15%)
Sentiment is the tone of the answers, from 0 (negative) through 50 (neutral) to 100 (positive), averaged over every answer. Words near your brand name count double.
Citation rate (15%)
Citation rate: for each answer, the share of its links that point to your site, averaged over every answer; an answer with no links counts as 0. Every platform can cite sources when it searches, ChatGPT and Claude included; an answer given from memory cites nothing. Links to profiles you register, such as your YouTube channel or app store page, count as your site.
How it works
- 1
Add the prompts your buyers ask
Write them the way a buyer types them, without your brand name. Your plan sets how many you can track: Free 5, Starter 25, Growth 50 and Pro 100.
- 2
We ask every platform in your plan
Free audits ChatGPT and Google AI Overviews; paid plans audit all 6. A platform that errors or runs out of time is left out of the score, not counted as a miss, and a prompt Google shows no AI Overview for is skipped on that column.
- 3
We check your site in the same run
robots.txt is read for 18 AI crawlers, and your page is requested as each crawler that fetches pages, beside an ordinary browser request. A crawler that robots.txt allows and your firewall refuses is reported, because nothing in your own browser would show you that block.
- 4
We read every answer
Each answer is checked for your name, its tone, its links, and any fact that contradicts your brand profile.
- 5
We score it
The four metrics are averaged with the weights above, then multiplied by 0.2 + 0.8 × coverage, where coverage is the market-weighted share of the platforms in your audit that mentioned you at least once. The coverage weights, our estimate of each platform's reach, are ChatGPT 36%, Google AI Overviews 18%, Perplexity 18%, Google AI Mode 10%, Gemini 9% and Claude 9%; each platform also gets its own score, without the multiplier. Mentioned on no platform, the score is 0.
- 6
You get the fix list
Fixes are ranked by priority, each with an effort estimate and the reason. Every audit is saved, so the next one shows what moved.
A score, worked through
10 prompts on all 6 platforms is 60 answers. Say 14 of them name the brand, all of them on ChatGPT and Perplexity.
Example output
- Mention rate
- 23% (14 of 60 answers)
- Accuracy
- 91
- Sentiment
- 62
- Citation rate
- 6%
- Weighted average
- 46.7
- Coverage
- 54%, the weight ChatGPT and Perplexity carry
- Multiplier
- 0.2 + 0.8 × 0.54 = 0.63
- Visibility score
- 30
Illustrative inputs; the arithmetic uses the same weights and multiplier as the product. Being absent from 4 of 6 platforms costs this brand 37% of its average.
A real audit, read as a funnel
On 15 October 2026 we audited Nike, Adidas, Puma and Under Armour with the product’s own scoring code. Five prompts a shopper types without naming a brand went to ChatGPT, Perplexity and Claude, so each brand was counted on the same 15 answers.
From a real audit
| Under Armour | ChatGPT | Perplexity | Claude | All three |
|---|---|---|---|---|
| How it answered | From memory | Web search | From memory | — |
| Answers naming it | 2 of 5 | 3 of 5 | 4 of 5 | 9 of 15 |
| Answers naming it first | 0 | 0 | 0 | 0 (Nike: 13) |
| Share of brand mentions | 6% | 13% | 12% | 8% |
| Links to underarmour.com | None given | 0 of 75 | None given | 0 |
Read down the last column. Of 15 answers, 9 named Under Armour, none named it first and none linked to its site; Nike was named first in 13.
One prompt carried the worst of it. On "best running shoes 2026" no platform named Under Armour, and its fix list opened with two critical items about that prompt.
Counted by the product’s own functions over the saved answers, on the models it used that day. Gemini's answers in that run were cut short by a token limit on our side and are left out, and Google AI Overviews was not part of it.
Read the AI visibility audit of Nike, Adidas, Puma and Under ArmourFix it with Agent V
Agent V is the assistant beside your dashboard. Ask it to explain your score and it reads your latest audit and walks you through what moved the number: which engines named you, where you were absent, and what that cost in coverage.
Ask it to compare two audits and it lines them up for you, so you can see whether a fix you shipped changed anything. It can also read a page on your own site as text, to check whether the facts an answer missed are on it.
Anything that changes your account, like running an audit or adding a keyword, comes to you as a card to confirm. Nothing changes until you do, and V never edits your site. It runs on Anthropic's Claude models and is on every plan.
Who it is for
- Founders and marketers who need to know what AI tells buyers before a buyer repeats it on a sales call.
- Teams running several brands: Growth covers 3 brands and Pro covers 10.
What it does not do
- It is not a ranking any engine publishes. The score is our summary of engines that work differently, so read the per-platform columns and the answers before the headline number.
- Your prompts are a sample of what buyers ask, not of what the engines search for. Google says AI Overviews and AI Mode may use a "query fan-out" technique, "issuing multiple related searches across subtopics and data sources", and no outside tool sees those searches.
- It calls each vendor's API, not the consumer app. ChatGPT and Claude are asked with web search on and decide per prompt whether to search; an answer from memory reflects training data, which ends in May 2026 for ChatGPT and June 2026 for Claude.
- It asks from one country: your first market. Your prompt carries no location; ChatGPT, Claude and Perplexity search from that country, AI Overviews and AI Mode are read from its Google results, and Gemini takes no location.
- It does not cover Microsoft Copilot or Meta AI.
- It sees answers, not clicks. Traffic from AI platforms is not part of the score.
Questions
How do I audit my brand’s visibility in AI search?
Ask each AI platform the questions a buyer types without your name, and record whether each answer names you, in what tone and with what links. Then check every fact in the answers against a primary source, and check that AI crawlers can reach your site. An audit runs those steps on every platform in your plan.
How often should I run an AI visibility audit?
Your plan sets the monthly allowance: Free 1, Starter 15, Growth 50 and Pro 150. On Starter and up you can schedule a weekly or monthly re-audit, which counts against the same allowance.
Should my prompts include my brand name?
Not for visibility. A prompt that names you guarantees a mention, so the prompts we suggest leave your name out, and the prompt composer warns that one which names you "measures how AI describes you, not whether it finds you." Keep a few branded prompts for the accuracy check: asking who runs the company is where we found a wrong CEO date in our sportswear audit.
Why did my score move when nothing on my site changed?
Engines can answer the same question differently from one run to the next, and vendors update their models. A platform that failed in one run drops out of that run instead of counting as zero.
Do I need to enter my brand facts?
Only for the accuracy check. Without facts there is nothing to compare an answer against, so accuracy stays at 100 and the audit marks it as not checked.
Can I run an AI visibility audit for free?
Yes. The Free plan gives you 5 prompts and 1 audit a month on ChatGPT and Google AI Overviews, plus the AI crawler access check and the fix list, with no card. Paid plans add Google AI Mode, Gemini, Perplexity and Claude and start with a 7-day trial that does take a card.
How do I choose an AI visibility tool?
Ask three things of any tool, ours included. Does it show the exact prompt and the full answer behind every number, does it say whether each answer searched the web, and does it check facts and crawler access or only count mentions?
Explore More Features
Hallucination Detection
AI states your facts with confidence. Some of them are wrong.
See how it worksMulti-Platform Coverage
ChatGPT, Google AI Overviews, Google AI Mode, Gemini, Perplexity and Claude. Measured one by one.
See how it worksPrescriptive Fixes
Not "consider improving". A ranked list, with the evidence for each item.
See how it worksSee What AI Sees
Start on the Free plan. No card needed.