01
Mention rate
Answers naming the brand ÷ total recorded answers
Whether the entity enters the answer at all. A mention is not automatically a recommendation.
Free tracker · 30-prompt baseline · no signup
Measure what AI answers mention, recommend, and cite.
A robots.txt allowance is not ChatGPT visibility. Record actual answer-engine results across a fixed prompt panel, preserve the source and context, and track mentions, recommendations, owned-site citations, competitors, and factual accuracy as separate rates.
Published July 23, 2026. The tracker does not call answer-engine APIs, scrape private account data, or transmit recorded observations to Spec Social. A result log is measurement evidence—not a ranking or citation guarantee.
Observed answers · separate rates
Record what the answer actually did.
This tool does not query AI providers or invent a universal visibility score. You run a fixed prompt, record the returned answer and citations, and compare the same rates over time. Everything stays in this browser unless you export it.
Mention rate
—
0 of 0 recorded answers
Recommendation rate
—
0 explicit recommendations or shortlist inclusions
Owned-site citation rate
—
0 answers cited at least one owned URL
Accurate-answer rate
—
Rate the entity facts when the brand is discussed
0 total runs0 surfaces covered0 unique owned URLs cited
Add one observation
One prompt. One returned answer.
Observation log
No answers recorded yet
Start with the downloadable prompt panel. Run prompts exactly as written, then record the returned brands and linked sources here.
The measurement model
Five observations. No fabricated “AI visibility score.”
These rates answer different questions and should remain separate. Combining them can make a frequently mentioned but inaccurately described brand look healthier than it is.
02
Recommendation rate
Answers explicitly recommending the brand ÷ total answers
Whether the brand makes a shortlist or direct recommendation—not merely background prose.
03
Owned-site citation rate
Answers citing an owned URL ÷ total answers
Whether the website is used as a linked source. Third-party citations should be recorded separately in notes.
04
Accurate-answer rate
Fully accurate answers ÷ answers with accuracy rated
Whether surfaced facts about location, services, people, pricing, and positioning match the current public record.
05
Competitor appearance frequency
Distinct answers naming each competitor
Which entities consistently occupy the answer space when the tracked brand is absent or included.
A repeatable protocol
Freeze the test before looking at the answer.
01
Define the entity and owned domain
Use one exact brand name and record the owned website separately from third-party sources. Resolve naming, location, service, leadership, and pricing facts before grading accuracy.
02
Freeze unbranded and branded prompts
Unbranded prompts measure discovery and shortlist share. Branded prompts measure entity recognition and factual integrity. Do not rewrite prompts after seeing the result.
03
Preserve answer provenance
Record the provider, model or mode, date, location context, citations, competing brands, and factual errors. Screenshots can support a private evidence archive, while the tracker stores structured observations.
04
Repeat on a fixed cadence
Run the same panel weekly or monthly. Compare rates and cited URLs over time, then connect changes to releases, indexing, earned mentions, corrected entity records, and newly published evidence.
Questions
What this tracker proves—and what it cannot.
What is AI search visibility?
AI search visibility is the observed frequency and treatment of a brand across answer surfaces for a defined prompt set. Useful measures include brand mention rate, explicit recommendation rate, owned-site citation rate, competitor frequency, and factual accuracy. A single universal score hides these different outcomes.
Does this tool automatically query ChatGPT or other AI providers?
No. It is an observation log. You run prompts in the answer surfaces you are authorized to use, then record the returned brands, citations, mode, location context, and accuracy. The tool stores those observations in your browser and does not imply API access it does not have.
Why are mentions, recommendations, and citations separate?
A brand can be named without being recommended, recommended without receiving a link, or cited as a source without being included in a commercial shortlist. Combining those events into one score makes diagnosis harder.
How many prompts should an AI visibility test use?
Use a frozen panel covering unbranded discovery, service-specific comparison, and branded entity checks. The downloadable baseline contains 30 prompts. Repeat the same panel on a fixed schedule and preserve provider, mode, date, and location context so changes are interpretable.
Can one ChatGPT answer prove AI visibility?
No. Answers can change with model, search mode, location, personalization, source freshness, and normal generation variation. One answer is a documented observation; repeated results across a frozen prompt panel provide stronger directional evidence.
Does allowing OAI-SearchBot guarantee ChatGPT citations?
No. Robots.txt permission only removes one access barrier for OpenAI's documented search crawler. It does not prove a successful crawl, index inclusion, citation, recommendation, answer accuracy, or appearance frequency.
Does Spec Social receive the answers entered here?
No. Brand names, prompts, citations, competitors, and notes remain in browser local storage. They leave the device only if the user exports and shares the CSV.