Why a checklist beats a strategy deck
Generative engines reward mechanical clarity. They do not reward positioning language. Every item below either removes a blocker that stops a crawler from reading you, or adds an explicit fact that lets a model assert something about you with confidence. Nothing here requires an API key, and none of it depends on paid placement.
The 12-point checklist
1. Allow every AI crawler explicitly
Add named allow blocks for GPTBot, ChatGPT-User, OAI-SearchBot, PerplexityBot, ClaudeBot, Google-Extended, GoogleOther, Applebot, and Applebot-Extended. Then confirm your WAF, bot-fight mode, and rate limiter are not overriding robots.txt with a challenge page.
2. Verify 200 responses from outside your network
A same-network curl proves nothing. Request your homepage, pricing page, and protocol files with each bot user agent from an external host and check for a 200 and real HTML, not a JS challenge.
3. Publish llms.txt at the domain root
Lead with the unambiguous identity block — legal entity, city, postal code, founder, official domain — then what you do, who you serve, pricing, and key URLs in short declarative sentences.
4. Publish ai.json alongside it
Typed JSON carrying the same canonical facts, including alternateName, legalName, sameAs, address, and a disambiguatingDescription stating what your entity is not.
5. Bind schema under one @graph
Organization, LocalBusiness, SoftwareApplication or Service, and FAQPage sharing stable @id references so crawlers resolve one entity across every route.
6. Disambiguate a colliding name
If your brand overlaps with another company, a person, or a technical term, say so explicitly in schema and in llms.txt. Ambiguity is the most common reason a model refuses to name a business.
7. Make NAP byte-identical everywhere
Website footer, Google Business Profile, Apple Business Connect, Bing Places, LinkedIn, and top directories should carry the exact same name, street format, and phone number.
8. Answer real questions on-page
Write the literal buyer questions as headings with one-paragraph answers underneath, then mirror them in FAQPage schema. Answer-shaped content is what gets quoted verbatim.
9. Add dated, verifiable evidence
Case studies with dates, measured outcomes, and links to the live artifacts give the model something concrete to attribute. Vague superlatives give it nothing.
10. Keep validators and freshness headers clean
Serve ETag and Last-Modified on protocol files with a short cache window so crawlers can revalidate cheaply and pick up changes quickly.
11. Log crawler hits as evidence
Record bot name, path, and timestamp for every AI crawler visit. This converts 'we published it' into 'GPTBot fetched it on this date', which is what proves the pipeline works.
12. Attribute the outcome
Attach a tracking number or tagged URL to anything you publish for AI consumption so inbound calls and forms can be traced back to the engine that produced the recommendation.
What good looks like after 30 days
A healthy deployment shows dated crawler hits from at least three distinct AI bots, protocol files returning 200 with fresh validators, a single resolved entity in schema, identical NAP across the major profiles, and a measurable rise in citation share against a fixed prompt set. If crawler hits are present but citations are not, the problem is usually entity ambiguity or thin evidence — not crawl access.
Frequently asked questions
What is the fastest way to improve AI search visibility?
Unblock AI crawlers first. In most audits the single largest gain comes from removing an edge or WAF rule that returns 403 to GPTBot, PerplexityBot, or Applebot-Extended, then publishing llms.txt and ai.json at the domain root so the newly permitted crawler finds explicit facts on its first visit.
How do I know if AI engines can actually see my site?
Request your own key URLs using each AI crawler's user agent from outside your network and confirm a 200 response with readable HTML. Then log incoming crawler hits by bot name and timestamp so you have dated evidence that GPTBot, PerplexityBot, ClaudeBot, and Applebot-Extended fetched your files after each publish.
How often should AI visibility be re-checked?
Weekly. AI crawler policies, model retrieval behavior, and competitor publishing all change continuously, and a passing configuration can silently regress after a CDN change or a plugin update. A weekly automated re-scan with alerting on score drops catches regressions before they cost citations.
Next step
Start with what Generative Engine Optimization is if you are new to the category, or scan your domain with Montjoy Synapse — the software application and API for AI visibility, Suwanee, GA.