Claude rewards something most organization websites are bad at: stating plain facts in plain sentences. This guide is a one-hour check of how Claude describes and recommends your organization — including making Claude read your own About page and tell you what's missing.
By Matt·Updated July 12, 2026
Quick answer
How do you check your organization in Claude?
Open a fresh Claude chat and ask the real questions your audience asks — about your topic, your services, and your organization by name. Ask once without web search to see what Claude already believes, then again with search on, and grade each answer for presence, accuracy, and citations. Then run the reverse test: paste your own About page in and see whether Claude can say what you do, for whom, from your words alone.
Claude builds answers from its training data plus live web search, and — as of July 2026 — it responds especially well to pages that state facts in clear prose, which makes your own site's clarity the main thing to audit here. (Background: how Claude finds information.) This page covers one engine; the full AI visibility audit covers them all.
Why check Claude specifically
Where ChatGPT's answers lean on what directories and reference sites say about you, Claude puts unusual weight on what your own pages say — read literally. A program page that opens with "empowering our community to thrive" gives Claude nothing to work with; one that opens with "we run a free after-school tutoring program for grades 3–8 at three locations in [county]" gives it everything. A Claude check is the fastest way to learn whether your website actually says what your organization does.
The stakes are the same as everywhere in AI search: when a parent asks Claude "where can my kid get free tutoring near [town]?", the answer they read may be the only impression you get. But the fixes a Claude check surfaces are unusually cheap — mostly rewriting first paragraphs, not rebuilding your web presence.
Three Claude behaviors this check accounts for
It reads you literally. Marketing language doesn't translate. Taglines, mission-speak, and "solutions" produce vague summaries; dates, places, services, and prices produce accurate ones.
It admits uncertainty. Claude will often hedge or say what it doesn't know rather than invent details. Every "appears to" and "may still" in its answer is a pointer to thin or conflicting information you can go fix.
Two sources, two failure modes. Like its siblings, Claude answers from a training-data memory that can be stale or from a live web search. The audit tests both, because each fails differently.
How to run the check, step by step
Five steps, about an hour, ending in a rewrite list for your own pages.
1
Write your answer key
One sheet of the correct facts: exact name, locations, hours, current programs or services, who qualifies or who you serve, and current leadership. A county government office, for instance, should include which services need appointments — that's precisely the detail an AI answer gets wrong. Without an answer key, a fluent 80%-right response will sail past you.
2
Ask what Claude already believes — search off
In a fresh chat, ask your branded questions plainly and add: "answer from what you already know, and explicitly flag anything you're uncertain about." Claude is unusually willing to mark its own gaps, which turns this step into a map of what the wider web has (and hasn't) taught it about you. Note every hedge — each one is a lead.
3
Ask again with web search on
Repeat the same questions, this time telling Claude to search the web, and add your unbranded questions ("where can seniors get help with taxes in [county]?"). Record each answer the same way — mentioned or not, accurate or not, your site cited or not, who else was named — in the AI visibility scorecard. Pay attention to which of your pages get cited: those are the ones already written plainly enough to quote.
4
Make Claude read your own pages
Paste the full text of your About page (then your key program pages) into Claude and ask: "Based only on this text, what does this organization do, who is it for, what does it cost, and how would someone start? List anything you can't answer." This isn't a visibility test — it's a clarity test. If Claude can't extract the basics from your own words, no engine can. Its list of unanswerables is your rewrite list; improving your About page shows how to work through it.
5
Confirm Claude can reach you at all
If search-on answers still miss you, check the plumbing: your site's robots.txt file (the short file that tells automated visitors which pages they may read) may be blocking Anthropic's crawler, ClaudeBot — sometimes left over from a blanket "block all bots" decision. And facts locked inside PDFs, images, or script-driven calendars are invisible to it. Can AI tools see my website? covers the diagnosis, and the website readiness check makes it systematic.
Diagnostic prompts to run
Copy, paste, and fill in the brackets. Run each in a fresh chat.
Baseline knowledge & honesty check (search off)
Prompt
Without searching the web, tell me what you know about
[organization name] in [city, state] — what it does, who it
serves, and its current leadership. Explicitly flag anything
you are uncertain about or don't know.
Recommendation test (search on, unbranded)
Prompt
Search the web: I'm looking for [the service you provide] in
[county or city]. Which organizations would you recommend, and
why? Cite the specific pages you relied on.
Plain-reader test (paste your own page)
Prompt
Here is the full text of a page from our website:
[paste your About or program page]
Based only on this text, explain what this organization does,
who it is for, what it costs, and how someone would get started.
Then list every basic question this page fails to answer.
What the responses tell you
Four common patterns, and what each one means.
Accurate, confident, and your pages cited. The healthy result. Note the pages Claude quoted — they're your house style guide for everything else you publish.
Hedged everywhere. "Appears to be," "may offer," "as of my last information." Claude is telling you the public record about you is thin or contradictory. Chase each hedge to its cause — usually a fact that exists in a PDF, a stale listing, or nowhere at all. Finding missing organizational information is the systematic version of that chase.
Fluent but vague. Claude describes you warmly without ever saying what you do — a mirror held up to your own copy. If the plain-reader test in step 4 also comes back with a long "cannot answer" list, the problem isn't your visibility; it's that your pages never state the basics.
Absent or misidentified. On unbranded questions Claude recommends others, or blends you with a similarly named organization. Check crawler access first (step 5), then how you compare to the organizations it did name — analyzing your competitors shows how. And because engines fail differently, run the same questions through the ChatGPT check, where third-party listings matter more, and the Gemini check, where your Google presence does.
What this check turns up in practice
Two representative scenarios — one vagueness problem, one identity problem.
The historical society Claude couldn't pin down
What the check found
A local historical society asks Claude what it does and gets warm fog: "an organization dedicated to preserving regional heritage." Asked directly whether the society runs a museum, offers school tours, or holds an archive, Claude says it can't tell. The plain-reader test explains why — the About page is three paragraphs of lovely storytelling ("preserving the stories that make our valley home") that never mention the museum, its hours, the research archive, or the tour program.
What it changed
The society keeps the storytelling but adds a first paragraph of plain facts: founded 1958, operates a free museum open Thursday–Saturday, maintains a 12,000-item research archive, offers school tours by appointment. On rechecks, Claude describes all of it correctly and starts surfacing the museum for "things to do with kids in [county]" questions.
The lesson: nothing was wrong with the society's visibility — Claude had found the page. The page just never said anything checkable. One paragraph fixed it.
The college that kept becoming a different college
What the check found
A small liberal-arts college shares most of its name with a larger university two states away. Claude's search-off answer blends the two — right town, wrong athletics conference, a graduate program the college doesn't offer. Search-on answers still import the university's details, because the college's own site says "the College" throughout and rarely states its full name and state together.
What it changed
The college adds one disambiguating sentence to its About page and footer — full name, city and state, "an independent college not affiliated with [university]" — and switches its directory listings and news boilerplate to the full name. Rechecks show Claude keeping the two institutions apart and hedging far less.
The lesson: AI tools can only tell similar names apart if the difference is written down somewhere. A single plain sentence, repeated consistently, is the whole fix.
Run your Claude check this week
Five items from zero to a rewrite list.
If you'd rather have this check — and the rewriting it usually leads to — handled for you, tell me about your situation.
Common questions
Claude doesn't know about our recent name change or reorganization. What do we do?
Claude's built-in knowledge lags the real world, sometimes by a year or more, so recent changes often aren't in it. You can't edit that memory directly. What you can do is make sure any answer built from a live web search comes out right: state the change plainly on your About page ("Formerly known as…"), update your major directory listings, and keep both names findable for a while. Search-backed answers pick up corrections within weeks.
Does writing plainly for Claude hurt our Google rankings or make our site boring?
No. Clear factual writing — who you are, what you do, who you serve, stated in the first paragraph — helps traditional search engines just as much as AI tools, and your human visitors most of all. Plain facts and warm storytelling aren't enemies; the fix is leading with the facts, then telling the story, rather than making a reader (or an AI tool) dig for basics.
Why does Claude sound less certain about us than ChatGPT does?
Claude tends to say so when its information is thin or conflicting rather than papering over it. Treat that hedging as free diagnostic data: an answer full of "appears to" and "may" usually means the public record about you is sparse, outdated, or contradictory. Fix the underlying pages and the hedging fades.
Claude confused us with another organization that has a similar name. How do we fix that?
Add a plain disambiguating sentence to your About page — your full name, your location, and if needed an explicit "not affiliated with…" — and use your full, consistent name on directory listings and partner pages instead of shorthand. AI tools can only tell similarly named organizations apart if somewhere on the web the difference is stated in words.