Evidence infrastructure for AI access

The independent record of what AI takes from your site.

perseus verifies every GPTBot, ClaudeBot, and PerplexityBot request against the crawler's own published IP ranges, hash-chains what's verified, and anchors the daily root outside our own systems. Publishers get in-forgeable proof of what was taken, against their policy. Everyone else gets a record perseus itself can't quietly rewrite.

Get started Start the free readiness scan →
perseus scan · AI readiness
Can AI crawlers actually reach your site?
Most scanners only check whether AI crawlers can reach your site. This one also checks whether they can read it — your robots.txt against 15 named AI crawlers, plus the raw HTML the way GPTBot and ClaudeBot actually see it, no JavaScript execution, alongside structured data and freshness. Free, no signup.
AI crawlability—
AI readability—
Structure & schema—
Citability & freshness—
Technical—
Overall readiness—/100

Where perseus sits

One neutral layer between your content and everyone who wants to act on it.

perseus sits at the boundary between your content and the AI companies, not inside either one. It verifies what actually crossed, and hands the record to whoever needs to act on it.

perseus doesn't price or resell the access it measures, and it can't rewrite the record after the fact. That's the same neutrality a Big Four auditor or Verisign trades on.

Why it matters

Google Analytics counts humans. The crawlers feeding your pages to AI never show up there.

AI crawlers don't run JavaScript, so they never load an analytics tag. In GA4, a day when GPTBot reads forty of your pages looks the same as a day when nothing happened at all.

Happening now

The data is already sitting in your server logs.

Every crawler request is a real HTTP hit your server writes down. Reading it means parsing raw logs across a dozen crawlers, each with its own IP ranges that change without notice. That's the tedious part we handle.

Google Analytics · today
"0 sessions from GPTBot, ClaudeBot, or PerplexityBot." They don't run JS, so GA never sees them.
perseus · today
GPTBot: 850 requests, verified. PerplexityBot: 420 requests, verified. ClaudeBot: 310 requests, declared.
What perseus does

Every number here comes from a request you can look up.

Nothing on the dashboard is modeled or estimated. Each hit is a request your server received, matched to the IP ranges the crawler's operator publishes. Pull your own logs and check any of it.

01

Every crawler hit, logged

Each GPTBot, ClaudeBot, and PerplexityBot request that reaches your server, attributed to the crawler behind it, straight from your server-side logs.

02

The real crawler, or something wearing its name

Anyone can put "GPTBot" in a request header. We check the IP against OpenAI's published ranges, so a scraper hiding behind the name doesn't get counted as the real thing. You see it flagged instead.

03

Key-page coverage

Did GPTBot fetch your pricing page, or just your homepage? See exactly which of your key pages each AI crawler reached, and which it skipped.

04

Coverage comes before citations

If a crawler never reads the page, no model can cite it. Coverage is the first thing worth fixing, well before you start worrying about how you rank inside the answer.

The dashboard

What shows up once your logs are flowing.

One screen: how many AI crawlers hit you, how many were the real thing, which of your pages they reached, and what to change. Below is a week of logs from a demo site.

app.perseus.io / acme.com
perseus dashboard: 1,641 AI crawler hits on acme.com, 1,430 verified against official IP ranges, 11 spoofed, and a list of recommended actions.
Verified vs. spoofed

Every hit checked against the source

Each request is matched to the IP ranges the crawler's operator publishes. Here, 1,430 checked out and 11 carried a crawler's name from an address that wasn't theirs.

Coverage

Which pages AI actually reached

Key-page coverage shows what each crawler fetched and what it skipped. A page no crawler reads is a page no model can cite.

What to do

The next move, spelled out

The evidence turns into specific actions: block the scraper at your edge, open up a page answer engines are missing, or decide what training crawlers get to keep.

Pricing

One site or a whole portfolio.

We're in private beta with a first group of sites and agencies, and we set pricing up with each of them. Book a demo and we'll get your logs flowing in.

Starter
For one site
  • Every AI crawler hit, verified vs. spoofed
  • Key-page coverage tracking
  • 7-day trend history
  • Spoof alerts
Join the waitlist
Growth
For teams that ship content
  • Everything in Starter
  • Unlimited key pages
  • 90-day trend history
  • Alerts on new spoofed traffic
  • Shareable reports
Join the waitlist
Agency
For multiple client sites
  • Many sites, one workspace
  • White-label reports
  • Webhook & API access
  • Assisted onboarding
Talk to us
Enterprise
Contact sales
  • Log-drains (Cloudflare Logpush, Vercel)
  • SSO / SCIM
  • Audit log & RBAC
  • Evidence dossiers on demand
  • SLA
Contact sales
Founding beta

Get on the list before pricing opens up.

We're onboarding founding customers by hand. Leave your email and your site, and we'll reach out when there's room.

See who's been reading you.

Create an account, point a Cloudflare Worker (or an access log) at your site, and you'll see the verified crawler hits, anything spoofing its way in, and which of your key pages GPTBot, ClaudeBot, and PerplexityBot have actually reached.

Get started