What the audit report actually looks like
Below is an anonymized excerpt of a full AI Visibility Audit report — a fictional "Example Plumbing AS" with numbers typical of our benchmark. Every claim in a real report works exactly like this: a check, a result, and the evidence for it.
1. The score — and what it means against your market
Methodology v0.3 (how the score works); the industry comparison comes from our own benchmark, computed with the same tests.
2. Category breakdown
| Category | Points | Reading |
|---|---|---|
| A. Accessibility for AI crawlers | 24 / 26 | Good — except the firewall (see A8 below) |
| B. Content without JavaScript | 9 / 15 | Homepage readable; service pages nearly empty without JS |
| C. Structured data | 4 / 15 | JSON-LD exists but incomplete; no FAQ markup |
| D. Content structure | 6 / 15 | Decorative headings, no question-shaped content |
| E. Entity clarity | 7.5 / 10 | E-mail and org.nr present; no phone; name inconsistent |
| F. Technical foundation | 5 / 12 | No sitemap; slow first response |
| G. Actual AI visibility (live tests) | 3 / 8 | AI knows the company — but recommends competitors |
3. Check results — with evidence
The full report lists all 33 checks this way: what was tested, the result, and the evidence — so you (or your web agency) can verify every line. A sample:
| Check | Result | Evidence |
|---|---|---|
| A2. GPTBot allowed in robots.txt | ✓ pass | No rule matches / for GPTBot |
| A8. Firewall does not block AI bots | 2.25 / 3 | GPTBot → HTTP 403; OAI-SearchBot, ClaudeBot, PerplexityBot → 200. The WAF blocks OpenAI's training crawler — robots.txt allows it, the firewall overrules it. |
| B1. Meaningful text in raw HTML (homepage) | ✓ pass | 2 340 characters of visible text before JavaScript |
| B4. Subpage content in raw HTML | ✗ fail | 2 of 3 sampled service pages serve <400 characters without JS |
| C2. Organization JSON-LD with name + address | 2 / 5 | Type present; address and url missing |
| C3. FAQPage structured data | ✗ fail | Not found on any audited page |
| E3. Organisation number in content | ✓ pass | Org.nr 912 345 678 found in the footer |
| F4. Server response time | 1 / 3 | TTFB 2.1 s — under 3 s, above the 1.5 s target |
| …plus the remaining 25 checks, each with points, evidence and source | ||
4. What AI actually answers — documented
The report ends the diagnosis with live tests: three fixed questions asked in ChatGPT, Perplexity and Gemini, every answer screenshotted and attached as an appendix. For this sample company:
Does AI know the company? ✓ Pass — ChatGPT and Gemini correctly describe what it does and where (2 of 3 engines required).
Does AI recommend it in its category? ✗ Fail — asked for "a plumber in [city]", all three engines recommend competitors; the company is never mentioned. The report names which competitors, with screenshots.
Does AI cite the company's website? ✗ Fail — one engine of three cites the domain; the threshold is two.
This is usually the section that lands hardest: not a technical score, but your actual absence from the answers your customers read.
5. The prioritized fix list
Every finding becomes a concrete, ordered task — highest recovered points per unit of work first. The top of this sample's list:
1. Unblock GPTBot at the firewall (recovers 0.75 pts, protects long-term AI knowledge of the company) — one WAF rule change; we provide the exact user-agent list.
2. Complete the Organization JSON-LD and add FAQPage markup (recovers ~6 pts, directly improves how AI understands the business) — we provide ready-to-paste JSON-LD.
3. Render service pages' content in HTML (recovers ~5 pts) — server-side or static rendering for the three template pages.
The full list covers every failed check, and the fixed-price implementation offer is scoped exactly to it — no "phase two" surprises.
Want this for your website?
The full audit costs 4 900 NOK — fully deducted if you order the implementation. Start with the free mini-audit: 5 of the 33 checks, your preliminary score in under a minute.
Get a free mini-audit → How we measure