HeardOf

Your website has a guest list.

It’s called robots.txt. Plenty of sites turn away the AI assistants their buyers ask, usually by accident. See who gets in at yours.

Free and instant. We read robots.txt and the home page; no AI engine is asked anything.

Guest list
lumenpayroll.com/robots.txt
Example from our sample audit
  • ChatGPT search
  • Google Search and AI Overviews
  • Perplexity answers
  • Claude search
  • Bing and Microsoft Copilot
  • Siri and Spotlight
  • Meta AI
  • ChatGPT, when a buyer pastes your link
  • Claude, when a buyer asks about your page
  • Perplexity, on request
  • Alexa answers
  • Le Chat by Mistral
  • DuckDuckGo answers
… of 13 turned away

Point at or tap a name to see what turning it away costs.

Training-only crawlers: 3 of 4 turned away. That’s your call; it doesn’t change answers.

One report. Every way you get found.

AI visibility, answer readiness, domain trust, consent and experience, each measured, scored and shipped with the fix.

  • AI visibility
    35%

    of 256 Claude answers name you (30–41%)

  • Domain trust
    57

    No DMARC record; SPF missing

  • Consent
    3

    trackers fire before anyone clicks

  • Experience
    61

    LCP 4.6 s on a phone

Getting in is half of it.

Once an assistant can read you, the question is whether it recommends you. Ask once and you get an anecdote. We ask your buyers’ questions eight times on each engine and count, with the margin of error.

Read a sample report
A buyer asks“What is the best payroll software for a 30-person startup in Germany?”
ChatGPT
Claude
Perplexity
Gemini
ChatGPT, answer 1 of 8

Asking four enginesAsking each engine 8 times…

 

How often Claude names you

Your buyers' questions, asked many times, counted with the margin of error. Not one screenshot of one answer.

  • Share of answers that name you, with a 95% interval
  • Who gets named instead, and the sources Claude cites
  • Re-measured after fixes; noise is labelled noise

“Best payroll software for a 30-person startup in Germany?” asked 256 times

Orbit Pay57%
Lumen Payroll (you)35%
Quill HR32%
Meridian25%

Bars are the share of answers that name each brand; the light band is the 95% interval.

Whether AI can read you

robots.txt per AI agent, structured data, and whether the answer sits where engines look for it.

  • Every AI crawler, allowed or turned away
  • Schema, titles and answer-first openings
  • Paste-ready robots.txt and JSON-LD
  • Claude-SearchBot and Claude-User may crawlpasses
  • OAI-SearchBot is blocked in robots.txtfails
  • No Organization schema on the home pagefails
  • Pages answer the question in the first paragraphpasses
  • Pricing page renders only with JavaScriptfails

Whether your domain is trusted

HTTPS, security headers and email authentication, checked the way a browser and a mail server see them.

  • Scored like MDN HTTP Observatory
  • SPF, DKIM, DMARC, MTA-STS with the exact record
  • JavaScript libraries with known vulnerabilities
  • SPFfail, no record
  • DKIMpass, google selector
  • DMARCfail, missing

Google, Yahoo and Microsoft reject or junk bulk mail without SPF, DKIM and DMARC. Each fix is the exact DNS record to add.

Whether “Reject” means reject

We load your home page four ways and record every tracker: before a choice, after reject, after accept, and with Global Privacy Control.

  • Trackers and cookies with their timings
  • Google Consent Mode signals
  • Observations, not legal advice
  • 0.8 sanalytics.ahrefs.com
  • 1.3 spagead2.googlesyndication.com
  • 2.1 sVisitor clicks “Reject all”
  • 2.9 sconnect.facebook.netafter reject
  • 3.4 s_fbp cookie set, 90 daysafter reject

How it feels on a phone

Core Web Vitals, accessibility, tap targets and overlays, measured in a real browser, then reviewed screen by screen.

  • axe-core and markup checks
  • INP from real taps, LCP and CLS
  • A heuristic review you can accept or dismiss
  • 3 accessibility rules fail
  • Tap targets under 24 px: 6
  • Largest Contentful Paint: 4.6 s
  • Lab INP: 180 ms
  • A modal covers 78% of the screen

Ask Claude. We bring proof.

Add HeardOf as a connector. Our tools fetch, measure and score; your own Claude does the reasoning, so nothing runs on someone else’s AI bill.

Connector URL
https://heardof.site/api/mcp

Claude: Settings → Connectors → Add custom connector. Claude Code and Cursor can run the local package for screenshots and consent scans.

  1. Is acme.com ready for AI search? Audit it and give me the top fixes.

  2. check_ai_crawlersOAI-SearchBot and PerplexityBot are turned away by robots.txt
  3. check_domain_trust64/100 · no DMARC record · HSTS off · jQuery 1.12.4 has 4 known vulnerabilities

  4. Three fixes first: let OAI-SearchBot and PerplexityBot in (two robots.txt lines), publish a DMARC record (p=none to start, the TXT record is below), and turn on HSTS. Each one comes with the evidence and the exact change.

Check

Who your site lets in, then how often each assistant names you when buyers ask. Every rate comes with its margin of error and the number of answers behind it.

Fix

A paste-ready change for every finding: robots.txt lines, structured data, rewritten page openings, and the review sites and sources the engines cite instead of you.

Re-check

A few weeks after the fixes ship, we measure again. A change is only called a change when the before and after ranges don't overlap; the rest is labelled noise.

Now get on ours.

We open a few audits at a time. Write yourself in and we’ll email you when it’s your turn.

  1. You write yourself in

    Email, site, company, role. Nothing else.

  2. We email when your slot opens

    Audits run a few at a time, so each one is read by a person.

  3. We agree your buyers' questions, then run it

    You get the report, the fixes, and a re-check after they ship.

Our guest list
HeardOf audits, opening a few at a time
Names stay private
You are
No newsletter. One email when your slot opens.