Docs · Getting Started
What Web2AI Does
Web2AI checks a website the way an AI system does, then tells you in plain English what it found. It works in three layers, in this order, so every claim is traceable.
Read the site live
Homepage, robots.txt, sitemap, llms.txt and up to 10 pages. Follows redirects, including meta refresh.
41 deterministic rules
Nine categories, 100 points. The same input always gives the same score.
Claude writes the report
From the findings only. It never sees your page, so it cannot invent a measurement.
Run Your First Scan
Enter your website address and the email that should receive the report, then run the scan. No password, no signup, three scans per day per visitor. The result page updates itself while the scan runs and the full PDF arrives by email.
Understanding Your Score
Scoring is additive. Each of the 41 checks earns points out of its maximum, partial credit included, and the nine categories add up to 100.
Grades
| Grade | Score | What It Means |
|---|---|---|
| A | 90 to 100 | Healthy. AI can reach, read and identify the site. |
| B | 80 to 89 | Good. A few signals missing. |
| C | 70 to 79 | Needs work. Reachable but hard to identify. |
| D | 60 to 69 | Weak. Several categories failing. |
| F | Below 60 | AI is likely skipping this site. |
The 41 Checks
Nine categories, each with its own point budget.
| Category | Checks | Points |
|---|---|---|
| Crawler And Index Access | 4 | 18 |
| Discoverability | 4 | 12 |
| Metadata And Social | 6 | 14 |
| Structured Data | 3 | 12 |
| Content Structure | 5 | 12 |
| Accessibility | 5 | 12 |
| Performance | 6 | 10 |
| Technical Health | 4 | 5 |
| Conversion | 4 | 5 |
Fix Files
Paid tiers ship upload-ready files: robots.txt, llms.txt, llms-full.txt, structured-data.html, meta-tags.html and faq-schema.html. Download them one by one from your report portal, or take the whole set as a ZIP. Emails carry the link to that portal rather than attachments, because mail filters block archives. Each file is generated from your own findings, so you upload rather than rewrite.
Fix Files: How To Upload
Every file in the ZIP is generated from your own findings, so the work is uploading rather than writing. Here is what each file is, where it goes and how to check it landed.
llms.txt And llms-full.txt
Plain text summaries of your site written for language models. llms.txt is the short index. llms-full.txt is the long form with more of your content in it.
Upload both to the root of your site, the same folder your home page is served from, so they answer at yourdomain.com/llms.txt and yourdomain.com/llms-full.txt. To verify, open those two addresses in a browser. You should see plain text, not your normal page and not a 404.
On WordPress, upload them over SFTP into the folder that holds wp-config.php, or use a file manager plugin if you do not have SFTP access. On Netlify or Vercel, put them in the folder your build publishes untouched, which is public in Next.js and Vite, then deploy.
robots.txt
This one replaces or merges with the file you already have. If you have no robots.txt, upload ours as it is. If you do have one, keep every Disallow rule that is already there and copy in our crawler and sitemap lines. Deleting your existing rules is the one mistake that matters here. It also goes in the site root, and you verify it the same way, by opening yourdomain.com/robots.txt.
structured-data.html
JSON-LD blocks that state what your business is. Paste the contents into the head of your page, or into the header scripts field your theme or site builder provides. Afterwards, run the page through the Schema Markup Validator at validator.schema.org and confirm it reads your Organization and WebSite blocks with no errors.
meta-tags.html
Title, description and Open Graph tags, also for the head. Replace the tags you already have rather than adding a second copy, because two title tags or two descriptions are worse than one.
faq-schema.html
FAQPage markup built from questions your site already answers. Fill in every TODO answer with real text first, then paste it into the head like the other two. Publishing it with the placeholders still in place tells AI systems your answers are TODO, so this is the one file to edit before it goes live.
AI Council
Several frontier models each score every issue for impact and assign a priority. The portal shows the average impact, the agreed priority and how many models agreed, so you can see where they differ rather than trusting one opinion.
Monitor
Monitor re-runs the same 41 checks on the same domain once a month, on its own, and emails you the result. The email leads with the new score and the change since last time, then lists which checks were fixed and which regressed, so a drop always comes with the reason attached.
To pause or cancel, open the Stripe customer portal. The link is in your report portal and in every Monitor email, and it is where the subscription itself lives, so you do not need us to stop it. Cancelling stops the next re-scan and leaves every report you already have readable.
Billing And Refunds
Payments run through Stripe. Prices are in USD. All sales are final and no refunds are provided. If a purchase does not deliver what this page describes, contact support and we will put it right. The full terms are on the Legal page.
Privacy And Data
Scans run on infrastructure SPARQIO owns. Your email is used to deliver the report. Marketing contact needs your explicit opt-in and every email carries an unsubscribe link. IP addresses are stored only as a salted hash and a deletion request wipes the contact, its audits and its hashed IP. The Legal page has the full notice.