The AI visibility checklist: 12 checks before you pay anyone
These 12 checks decide whether ChatGPT, Claude, Perplexity and Gemini can find you, read you and recommend you. Every one is checkable today, most in under five minutes. Check 1 caught our own site.
Access: can AI reach you at all?
1AI crawlers get a 200, not a 403
Firewalls, CDNs and bot protection block AI crawlers by default. If GPTBot or ClaudeBot gets a 403, you do not exist to that engine, whatever your content says.
How to check: ask your developer to request your homepage with the user agents GPTBot, ClaudeBot, PerplexityBot and ChatGPT-User and confirm each returns 200. On Cloudflare, review Security settings for "Block AI bots" and AI Crawl Control. This is the check that caught us.
2robots.txt welcomes answer engines
One managed setting can quietly write Disallow rules for every AI bot into your robots.txt.
How to check: open yoursite.com/robots.txt. Look for Disallow lines under GPTBot, ClaudeBot, CCBot, Google-Extended or PerplexityBot. If they are there and you did not choose them, remove them.
3A real sitemap.xml, listed in robots.txt
The sitemap is still how crawlers of every kind discover your pages fast.
How to check: open yoursite.com/sitemap.xml. Every page you care about should be in it, and robots.txt should point to it.
4Content readable without JavaScript
Many AI fetchers read the raw HTML. If your pages render as an empty shell until JavaScript runs, engines may see nothing.
How to check: view source on your key pages. If your actual sentences are not in the HTML, fix rendering.
Trust: can AI verify who you are?
5Schema.org markup on the money pages
Organization, Service or Product, and FAQPage markup state your facts in the format machines trust most.
How to check: paste your URL into a schema validator (validator.schema.org). You want Organization plus FAQPage at minimum.
6An llms.txt file
A plain-text summary of who you are, what you sell and where your key pages live, written for language models. Young standard, costs ten minutes, and the engines that read it get your story straight from you.
How to check: open yoursite.com/llms.txt. Ours is at showuplabs.com/llms.txt as a template.
7One consistent identity everywhere
Same business name, same one-line description, same offer on your site, LinkedIn, directories and socials. Engines cross-check; inconsistency reads as uncertainty, and uncertain facts get left out of answers.
How to check: read your own homepage, LinkedIn and top directory listing side by side. Would a stranger say they describe the same company?
Content: is there anything worth quoting?
8A two-sentence answer to "what do you do"
Engines quote clean statements. If your homepage cannot answer "what does this company do, for whom, at what price" in two sentences, an engine cannot say it either.
How to check: read your hero section aloud. Strip the adjectives. Is a concrete claim left?
9Pages that answer real buyer questions
Definitions, comparisons, pricing pages, checklists. The questions your buyers actually type are the questions engines answer, and they cite the pages that answer them cleanly.
How to check: list the five questions buyers ask before choosing you. Does a page on your site answer each one directly, near the top?
10Mentions in sources engines cite
Ask an engine "best [your category]" and read which sources it cites: roundups, directories, review sites, communities. If none of those mention you, that is the gap, and it is the strongest single lever on this list.
How to check: run the question, open the citations, search each for your name.
Measurement: do you actually know?
11You have asked the engines yourself
Not once, and not just your brand name. The questions your buyers ask, put to the engines they use, with a count of who gets named.
How to check: do it manually across ChatGPT, Claude, Perplexity and Gemini, or run the free audit, which does it against a live engine and scores you in a minute.
12You re-measure on a schedule
AI answers move as models update and competitors publish. A score from March means little in July. Monthly is enough; a calendar reminder is enough.
How to check: if you cannot name your score and the date you measured it, this one is failing. We automated ours: the live history.
Get your score before you change anything
Run the free audit first so you have a baseline number. Fix, then re-run, and watch it move. That is the whole loop, and it is the exact one we run on ourselves.