What brand facts should I give an AI visibility tool?
99Visibility Research · Updated October 21, 2026 · 6 min read
Short answer
Give an AI visibility tool the facts you can verify and that buyers ask about: current pricing and plan names, founding year, headquarters and service locations, product and feature names, and ownership or parent company. A tool can only call an AI claim wrong if it has a matching fact on record. When a fact is missing, the claim is unverifiable, not false, so gaps in your fact sheet become gaps in your accuracy checks.
Key takeaways
- An accuracy check compares what an AI says against facts you supplied, so the tool is only as good as your fact list.
- A claim with no matching fact on record is unverifiable, which is different from being proven wrong.
- The highest-value facts are pricing, founding year, location, product names and ownership, because errors there mislead buyers directly.
- Ask any vendor how it scores unverifiable claims, because some tools count them the same as false ones.
- Publish the same facts in plain HTML on your own site, consistently, so engines and tools see one version.
Why the tool needs your facts first
An AI visibility tool checks accuracy by comparison. It reads what ChatGPT, Perplexity or another engine says about you, extracts the specific claims, and matches each one against brand facts you have put on record. If the AI says your starter plan costs $29 and your record says $49, the tool can flag a wrong price. If you never recorded a price, it has nothing to compare against.
That makes your fact list the ceiling on what the tool can catch. A thin list produces a clean-looking report that misses real errors. A complete list lets the tool separate three outcomes:
- Supported: the AI's claim matches your record.
- Contradicted: the AI's claim conflicts with your record. This is a real error.
- Unverifiable: the AI made a claim your record does not cover. Nobody can say yet whether it is right.
The practical move: before you judge any accuracy score, check how many facts sit behind it. A short brand fact sheet written once saves you from trusting a number built on nothing.
The facts worth supplying
Prioritise facts that are specific, stable enough to record, and costly if an AI gets them wrong. A buyer who reads a wrong price or a wrong owner may never visit your site to correct it.
| Fact | Why it matters | What to record |
|---|---|---|
| Pricing | The most common buyer question, and it changes | Plan names, prices, billing period, what each plan includes |
| Founding year | Signals maturity, easy for models to guess wrong | One year, stated once |
| Location | Decides whether you serve the buyer at all | Headquarters city and country, plus regions you serve |
| Products | Models often merge, rename or invent products | Current product and feature names, and retired ones marked as retired |
| Ownership | An error here can misrepresent who you are | Parent company, acquisitions, independent status |
After those, add facts buyers compare on: key integrations, certifications, and headcount if you publish it. Leave out anything you cannot verify yourself. A wrong fact in your own record turns a correct AI answer into a false alarm.
Write each fact as one plain sentence with a date you last checked it. That date tells you when the record itself has gone stale.
Unverifiable is not the same as wrong
The split between contradicted and unverifiable is where tools differ most. Research metrics do not settle it. FActScore, a factual-precision metric from Min et al. (EMNLP 2023), groups claims the source contradicts with claims it lacks evidence for, under one "not supported" label. That works for benchmarks. For a brand, it means an employee count you never recorded can be scored like a price you know is wrong.
Other work handles gaps differently. Vectara's hallucination leaderboard reports an answer rate alongside accuracy, so responses that never came back are not counted as clean ones. The same logic applies to your audits: an engine that errored has given you no evidence either way.
What this means for you:
- Ask the vendor whether unverifiable claims lower your score, and by how much.
- Read flagged claims one by one before acting. Some flags reflect a gap in your record, not a false statement. This is covered in more depth in whether every flagged AI claim is actually false.
- When a flag is unverifiable, the fix is often adding the fact to your record, not chasing the AI.
These questions belong on your list when choosing a brand accuracy monitoring platform.
Edge cases that skew accuracy checks
Answers from memory, not the web
Many AI answers come from the model's training data with no live search. OpenAI's model documentation, checked in August 2026, lists a February 16, 2026 knowledge cutoff for the GPT-5.6 family and August 31, 2025 for the Instant path used in ChatGPT. A price you changed after those dates can appear wrong in ungrounded answers no matter what your site says today. Nobody outside the vendor controls that.
Facts your own site contradicts
If your pricing page, about page and footer state different founding years, engines and tools both see conflicting sources. Fixing contradictions on your own site is fully in your control. The same problem appears when language versions carry different facts.
Facts hidden behind JavaScript
Vercel's crawler study found that ChatGPT and Claude crawlers fetch JavaScript files but do not execute them, so they cannot read client-side rendered content. A price that only loads after scripts run may never reach those engines. Put load-bearing facts in the initial HTML.
Facts that change often
Promotions and trial lengths change faster than most fact sheets. Record the standing offer, not the campaign, or update the record on the day the offer changes.
What to do next
- Write the fact sheet. Start with pricing, founding year, location, products and ownership. One sentence each, with a last-checked date.
- Audit your own site against it. Find every page that states each fact and make them agree. Check that each fact appears in the HTML before JavaScript runs.
- Load the facts into your tool. Then run an audit and sort the flags into contradicted and unverifiable.
- Fix the gaps first. Add missing facts to your record where buyers would care. Then deal with true contradictions using the steps in handling AI hallucinations about brands.
- Review monthly. Update the sheet whenever pricing, products or ownership change, then re-run the audit.
If you want a quick first look before writing the sheet, the free checker on this site shows what AI engines say about you for up to three buyer questions. It is a snapshot, but it shows which facts come up most. None of this guarantees an engine will repeat your facts. It makes sure that when one gets them wrong, you know.
Frequently asked questions
How many brand facts should I start with?
Start with the five that buyers ask about most: pricing, founding year, location, products and ownership. That is enough to catch the errors that cost you deals. Add integrations, certifications and headcount once the core list is accurate and your site agrees with it. A short list you keep current beats a long list with stale entries, because a stale fact in your record creates false alarms.
Will giving a tool my facts make AI engines describe me correctly?
No. The tool uses your facts to check answers, not to change them. What engines say depends on their training data, what they retrieve, and how the model writes the answer, and you control none of the last two directly. What you control is publishing the same facts clearly on pages engines can fetch. Stating a specific fact on a fetchable page gives an engine a sentence it can quote.
Should I include facts I would rather not publicise?
Only record facts you can verify and are willing to state. The record exists to judge AI claims, so if an AI mentions something sensitive, such as a past acquisition, you want to know whether it is right. If a fact is private, leave it out and accept that related claims will show as unverifiable. Never put a fact in your record that contradicts what you publish.
Why does my accuracy score change when I add more facts?
More facts give the tool more to compare. Claims that were unverifiable can become supported or contradicted, so the score can move either way. That is the score becoming more informative, not your AI coverage getting worse. Compare scores only when the fact list behind them is similar, and note the date you changed the list so you can explain any jump.
Sources
This page is AI-assisted. It was written from a verified evidence base and checked against our editorial policy. Engine behaviour changes often; facts are as of October 21, 2026.