Check whether AI systems can actually read, interpret, and recommend your site.
GEO Fix team5 min read
Most owners check if their website is in ChatGPT by typing their company name. Buyers don't search that way. They ask for categories.
This post is a testing methodology — prompts, a scorecard, and what to record. It is not a diagnosis of why you're missing (→ that post) and not a fix guide (→ implementation post.
| Test | Question it answers | How |
|---|---|---|
| Recommendation test | Does ChatGPT name or cite you for buyer queries? | Manual prompts (below) |
| Access test | Can AI search crawlers read your site at all? | robots.txt review + free Express Check |
You need both. ChatGPT can "know" your brand from old training data while failing to cite your site in a live search answer. Access test without recommendation test = you don't know if buyers see you. Recommendation test without access test = you can't explain why.
Pull from sales calls and proposal requests — not marketing brainstorms.
Template prompts:
Do not use: "Do you know [Company Name]?" — that measures brand recall, not category visibility.
Run each prompt 3 times on different days. ChatGPT answers shift. You're scoring patterns.
For each prompt run, record:
| Score | Meaning |
|---|---|
| 0 | Competitors named; you're absent; no URL to your domain |
| 1 | You're mentioned but not in the top shortlist; or description is wrong |
| 2 | You're in the shortlist; no URL citation |
| 3 | You're in the shortlist and your URL appears in sources |
Scorecard template:
| Prompt | Run 1 | Run 2 | Run 3 | Median |
|---|---|---|---|---|
| Best [category] in [city] | ||||
| Recommend [service] for [type] | ||||
| … |
Interpretation:
| Median across prompts | Likely bucket (see diagnosis post) |
|---|---|
| Mostly 0 | Crawler block or severe misclassification |
| Mix of 0–1 | Category/structure gap or authority gap |
| Mostly 2–3 | Visibility OK on tested queries — expand prompt set |
Median 0 across category prompts + strong Google rankings → run the access test next. Don't open a schema guide yet.
Check whether AI search crawlers can reach your site:
Red flags to watch for:
| Signal | Healthy | Problem |
|---|---|---|
| AI search crawlers reachable | Most or all allowed | None or one of several blocked |
| robots.txt | OAI-SearchBot allowed | Blocked or missing explicit allow |
| WAF / Cloudflare | No blanket "block AI" rule | Managed rule blocking search crawlers |
Recommendation test median 0 + blocked crawlers = Bucket 1 in our diagnosis framework. Fix is technical — Starter from $199 or Pro from $399 one-time covers most Bucket 1–2 deployments (see the fix guide), not more blog posts.
Pick one competitor ChatGPT cited in your scorecard. Don't audit their whole site — check three things only:
| Check | Competitor | You |
|---|---|---|
| Named in your median-0 prompts? | Yes | No |
| robots.txt allows OAI-SearchBot? | Usually yes | Compare |
| Homepage first paragraph states category plainly? | Usually yes | Compare |
This isn't competitive intelligence theater. It's one row of evidence for which bucket you're in.
Pick your highest-stakes category query. Run it in Google and ChatGPT same day.
| Channel | Record |
|---|---|
| Google organic top 5 | Who ranks? |
| Google AI Overview | Who's mentioned? |
| ChatGPT median score | From Step 2 |
Divergence is normal. Semrush found AI-referred visitors convert roughly 3x more often — so "we rank on Google" doesn't answer the ChatGPT column.
Keeping tests separate from fixes stops the "read six articles, same checklist" problem.
Use category prompts + the 0–3 rubric. Brand-name prompts don't measure buyer visibility.
ChatGPT has no site status tool. Express Check tests crawler access and structure — the access half of this methodology.
Monthly, or after any Cloudflare, robots.txt, or CMS security change.
No. Search Console is Google-only. Use this scorecard for ChatGPT.
Key takeaways
Find technical blockers, missing context, and weak AI-readiness signals in minutes.
No payment, no subscription.
