Get picked by AI assistants
Find out what AI assistants currently say about your product, then make your pages readable enough that they get it right.
This is for you if
- An assistant describes your product as something it isn't.
- You don't know whether AI crawlers are allowed on your site.
- You have no llms.txt and can't tell whether that matters.
What this will not doYou can make your product readable and get it described elsewhere. You can't make an assistant recommend you — that stays its call, and the models change without warning.
Your plan
Week 1
Step 1Mark what a machine gets wrong about your product
Run the Homepage Check, then go through the claims it pulled from your pages one at a time. Tick the ones that are right and cross out the ones that aren't.
Go deeper in the playbookDone whenYou've written down what a machine says your product does, and marked each part right or wrong.
Step 2Let the crawlers that cite you reach your site
The crawler that fetches a page to answer a question usually isn't the one collecting training data. Set a rule for each separately, and check your CDN as well as robots.txt.
Go deeper in the playbookDone whenThe citing crawlers return 200 on a product page, and anything blocked is blocked on purpose.
Step 3Publish an llms.txt that points at pages that exist
A short file listing your real pages in plain language. A file full of dead links is worse than no file, so validate it after you deploy.
Done whenYour llms.txt is served as plain text and every link in it returns 200.
40 minGo deeper in the playbookStep 4Put your product facts in the server response
Plenty of these fetches never run JavaScript. If your product name and description only show up after hydration, they might as well not be there.
Done whenName, description, and JSON-LD are all in the raw HTML before any script runs.
45 minGo deeper in the playbook
Week 2
Step 5Get one outside listing that matches your site
A claim that only appears on your own site has exactly one source. Submit the same product facts to one directory, then check the live listing against your homepage.
Go deeper in the playbookDone whenAt least one independent page describes your product without contradicting your homepage.
Ticked steps are saved in this browser only. Nothing is sent to a server.
Get your product found
The same work, broken into every individual check, with what commonly goes wrong at each one.