How to Read Your AI Visibility Scan Results

You ran a scan, the number is worse than you hoped, and the instinct is to look at the overall score and feel bad.

The overall score is the least useful thing in the report. Here is the order to actually read it in.

1. Start with the on-page signals, not the score

If the technical checks show a blocked AI crawler or content that only exists after JavaScript runs, stop reading the rest. Nothing else in the report is meaningful, because the assistants could not read your site at all.

Fix that first, wait two weeks, and re-scan. A large share of zero scores are explained entirely by these two items, and both are cheap fixes.

2. Read the per-assistant breakdown

This is the diagnostic section, and the pattern tells you where to work.

Strong on Gemini and AI Overviews, weak on ChatGPT and Claude. Your Google presence is decent and your off-site entity is thin. You exist in Google's data and almost nowhere else. Work on directory breadth, association and manufacturer listings, press, and community mentions.

Weak on Gemini and AI Overviews, better elsewhere. Unusual, and it points at your Google Business Profile. Check completeness — services, service areas, categories, Q&A, photos. Also check proximity: if you are far from the population center you are targeting, that is a real constraint.

Weak on Perplexity specifically. Perplexity mirrors search results and directory pages closely. Weakness here means you are missing from the sources it cites. Run your queries in Perplexity, read every citation, and find which directories and local roundups you are absent from.

Weak everywhere except branded questions. The most common pattern. The assistants know you exist when asked directly but never volunteer you. This is a discovery problem, not an identity problem — you need directory presence, reviews with specifics, and content that answers the questions people actually ask.

Zero everywhere, including branded. Either a technical block, or your entity is thin enough that the assistants have no confident picture of you at all. Check the technical items, then count how many independent sources mention your business name.

Low on Claude but fine elsewhere. Often normal. Claude names fewer specific businesses by design. Look at whether competitors are being named on Claude — if they are and you are not, that is a corroboration gap: licensing, certifications, and consistent facts across sources.

3. Read the competitor list

This is the most actionable section and most people skim it.

Every business named in your place is a live case study. For the three that appear most often:

1. Open their Google Business Profile. Count the services and service areas against yours.
2. Search their name in quotes. Count independent mentions against yours.
3. Read their last twenty reviews. How specific are they?
4. View source on their homepage. Do they have schema? Is their content in raw HTML?
5. Check whether they are on the local roundups you are missing.

Nine times out of ten you will find the difference in fifteen minutes, and it is usually mundane — more directory presence, more complete profile, more specific reviews.

Also note how many distinct businesses appear across all your questions. Three names repeating means a consolidated market that takes longer to break into. Twelve different names means a fragmented market where a complete profile can move you quickly.

4. Read the verbatim answers

Do not just count mentions. Read what was actually said.

When you are named, what reason is given? "Known for emergency response" or "family-owned since 1998" — that justification came from text somewhere public. Note which of your attributes are landing.

When competitors are named, what reason is given for them? That tells you what attribute the assistant considers relevant for this query, and whether you have a public claim on it.

What does the assistant recommend people look for? If it says "check for licensing and insurance," and your license number appears nowhere in plain text on your site, that is a direct instruction.

How does it describe your category? Sometimes the framing reveals a whole content gap — an assistant that keeps steering toward "get three quotes" is telling you that price transparency is the unmet need.

5. Read the branded question results

These are the questions where your business name is in the prompt: "is [business] any good," "tell me about [business]," "is [business] licensed."

Two things to look for.

Errors. Wrong phone number, old address, services you dropped, wrong hours, confusion with a similarly named company. Write down every one. Each error exists in a specific public source, and finding and fixing that source is usually the fastest win in the entire report.

Vagueness. If the assistant can only say "it appears to be a plumbing company in Tampa," your entity is thin. There is nothing specific enough about you in public to say more. That is a signal to publish specifics — years in business, credentials, specialties, real numbers.

"I don't have information about that business." The clearest possible statement that your entity does not exist. Directory breadth is your priority.

6. Then pick exactly one thing

The most common failure after a scan is opening twelve tabs and doing none of it.

Pick the single highest-impact item from what you read:

• Technical block → fix robots.txt or rendering. Today.
• Thin Google profile → block two hours this week.
• Thin entity → claim ten listings this month.
• Missing from Perplexity's citations → contact two roundup publishers.
• Factual errors → track down and fix each source.
• Generic reviews → change your review request text today.

Do that one thing. Re-scan in thirty days. Then pick the next one.

What the overall score is actually for

One purpose: comparison against your own previous score.

The absolute number means little on its own — it depends on market size, competition, and how consolidated your category is. A 22 in a competitive metro can represent a stronger position than a 45 in a small town.

What means something is 8 last month and 24 this month. Track the delta, ignore the absolute.

Re-scanning

Monthly is right. Same questions, same method, so the comparison is real.

Expect the assistants to move in a predictable order: Google AI Overviews and Gemini first, Perplexity next, ChatGPT and Claude last. If nothing moves anywhere after ninety days, go back and re-check the technical items — something you thought was fixed probably was not.

Run one now

Our free scan gives you all of the above — per-assistant grades, the competitor list, every verbatim answer, branded question results, and fifteen on-page signals.

About fifteen seconds, free, no account. Then read it in the order above.