The DIY AI visibility audit

Test 1 — Can machines read you? (10 minutes)

Open a terminal (or any online HTTP-fetch tool) and request your homepage the way crawlers do — raw, no browser:

curl -s https://yoursite.com

Now search the output for: your brand name, your main service phrase, your city, your title tag. All present? Pass. What you get instead — an empty div, a wall of scripts, or an endless redirect — means AI crawlers see nothing, and no other test below can score well until this is fixed. Check three pages: home, your top service page, one article. Score: 0 (invisible), 1 (partial), 2 (full content in raw HTML).

Test 2 — The money questions (15 minutes)

Write the ten questions a ready-to-buy customer asks in your category — "best [service] in [city]," "[product type] that actually works," "who should I hire for [job]." Ask all of them to ChatGPT, Perplexity, Gemini and Claude. Log in a sheet: were you named? who was? which sources got cited?

This is the scoreboard that matters — your maturity level, measured directly. The cited-sources column doubles as your citation-building target list, so don't skip it. Score: 0 (never named), 1 (named occasionally, one engine), 2 (named consistently, multiple engines).

Test 3 — What do they say about you? (5 minutes)

Ask each engine directly: "What is [your brand]?" and "Is [your brand] trustworthy / any good?"

You're checking two failure modes. Ignorance — "I don't have information about this brand" — means your public footprint is too thin to corroborate. Misrepresentation — wrong services, old prices, a stale description — means a trusted source somewhere holds bad data, and the engines learned it. Note every wrong fact; each one is traceable to a fixable source. Score: 0 (unknown), 1 (known but wrong/thin), 2 (known and accurate).

Test 4 — The consistency sweep (10 minutes)

Open your homepage, your map/business profile, your top two directories and your main social bio side by side. Compare: business description, services, phone, address, hours, even the spelling of your name. Engines cross-check these constantly; disagreement reads as risk, and risky brands get omitted from answers.

While you're there: does your homepage state who you serve, what you do, where, in one plain extractable sentence? That sentence is what answers get composed from. Score: 0 (contradictions everywhere), 1 (minor drift), 2 (identical story everywhere, canonical sentence present).

Scoring, and the one-fix rule (5 minutes)

Add it up — 8 points possible:

• 0–2: You're at LV0–1. Everything routes through Test 1's fix: get readable, then get describable. Nothing else pays yet.
• 3–5: LV2 territory. Plumbing works; now it's extractable answer-pages, consistency cleanup, and starting the publishing cadence.
• 6–8: LV3+. You're in answers — the game is coverage, citation building on the sources from Test 2, and monthly re-measurement to defend the ground.

The one-fix rule: the layers gate each other, so fix only the lowest-scoring test, then re-run the audit next quarter. Forty-five minutes, four times a year, beats every dashboard you'll never open. (Want the professional version — full crawl, citation map, edge fixes scoped? That one's free too.)