GEO audit example: from findings to an action plan
Before you commission a GEO audit, ask what your team will be able to do with the report. A useful finding lets someone open the affected page, reproduce the issue and later confirm that it was fixed. A score with none of that detail leaves most of the work unexplained.
This example comes from our review of geo-rank.ai on 7 September 2026. We examined public pages and source files. We had no Search Console, analytics or server logs for this review. Read it as a diagnostic example; it is not a case study of traffic or sales growth.
Follow-up: all 347 sitemap pages checked on 12 October 2026
We repeated the public technical check after the site had grown. Every one of the 347 URLs in the published sitemap returned HTTP 200 directly, had a self-referencing canonical and allowed indexing in its HTML. None returned an X-Robots-Tag restriction. The earlier indexing prohibition was no longer present on the eight translations of the Google rankings and AI citations guide.
| Check | Observed result | What it establishes |
|---|---|---|
| URLs in the sitemap | 347 unique URLs; 347 direct HTTP 200 responses | The listed pages were reachable from the test connection |
| Canonical and indexing directives | 347 matching canonicals; no noindex or X-Robots-Tag restrictions | No indexing prohibition was found on these pages |
| Published HTML versus source files | 347 exact matches | The tested content was already published |
| Internal discovery | Every sitemap page had an incoming HTML link from another page | The site contains crawlable routes to the pages |
| Nonexistent test URL | HTTP 404 | The tested missing page did not return a misleading HTTP 200 |
Download all 347 URL results (CSV) and the measurement notes. The snapshot was taken before that day’s content and navigation updates. Each row includes the returned canonical, indexing directives and a SHA-256 fingerprint of the HTML.
How to repeat a page check
Use GET so you receive the page body as well as the headers. For example:
curl -sS -L --max-redirs 5 --max-time 30 \
-D headers.txt -o page.html \
https://geo-rank.ai/en/geo-audit/Read the status and any Location or X-Robots-Tag headers in headers.txt. In page.html, check the canonical, robots directive, main text and links. Record the requested and final URL separately; a successful final response can hide a redirect.
What to investigate when the crawl passes
Technical access is only the first part of the diagnosis. Our internal-link check found country and language guides with just one incoming link from another page in the same language: the blog index. Links from the multilingual guide give readers a more useful route to those specific examples. We also found dates on blog cards that disagreed with the articles, and no separate analytics event for clicks on the contact email address.
Those are actionable findings, but they do not establish why Google omitted a page or an assistant chose another source. This review did not use Search Console, analytics reports or verified crawler logs. Requests with crawler names in the User-Agent header returned the expected HTML on three sample pages, but that does not test access from the crawlers’ actual networks.
Google says a page must be indexed and eligible for a snippet to support an AI Overview or AI Mode answer. Meeting its technical requirements does not guarantee selection. Use the Google AI features documentation to separate eligibility from performance. An audit should identify the next missing piece of evidence rather than declare success because every URL returns 200.
The September audit: define the scope before interpreting results
All 216 URLs in the sitemap returned HTTP 200: the server delivered every page. Eight language versions of one guide, though, also carried noindex, nofollow, which asks search engines to keep them out of the index. So the crawl confirmed that the pages loaded. It told us nothing about whether Google had indexed them.
A report should keep three statements apart: the page is accessible, indexing is permitted, the search engine has indexed it. Each needs its own evidence. Fold them into a single “AI visibility” score and the client loses the very distinction they need.
A reproducible technical finding
On /ru/privacy/, the relative path ../assets/site.css resolved to /ru/assets/site.css. That resource returned 404, leaving the page without its shared stylesheet. The same path assumption affected a script and navigation links across 16 legal documents.
| Report field | What the finding should say |
|---|---|
| Observation | The page opens but requests its stylesheet at the wrong URL |
| Evidence | Final page URL, resolved resource path and 404 response |
| Impact | Missing page styles and links that lead to the wrong pages |
| Fix | Use /assets/site.css and check the other resource paths |
| How to verify the fix | CSS and JS return 200; home and language links reach the intended pages |
These legal documents were excluded from indexing on purpose. The broken interface still mattered to visitors, but nothing justified claiming that it explained the site’s absence from an AI answer. Report the impact you observed and do not invent a ranking consequence.
Content findings need evidence too
The audit page offered a fixed scope without giving buyers enough detail about which pages, languages and checks it included. Reading the offer was enough to show the gap. No server error was involved.
The page needed to explain how we would agree on the URLs, page templates, languages, questions and repeat checks. It also needed a list of deliverables and any work charged separately. The test was practical: could a buyer compare the proposal with the final report and see whether the agreed work had been completed?
Assign each fix and explain how to check it
Start with problems that stop important pages from working. Then check whether the content gives buyers enough information to choose a service. For each task, name who is responsible, what they need before starting and how the result will be checked. An editor needs confirmed service terms. A developer needs a list of affected templates.
A red warning in an audit tool does not tell you which problem to fix first. A missing optional file and an inaccessible service page deserve different treatment. Google’s own AI search guidance explains that special AI files and markup are not required for its features. That statement covers Google only, so an audit should name the systems and conditions it actually tested.
What to ask before paying
Ask for one complete, anonymized finding and how it was verified. Find out whether implementation is included and who checks it. If monitoring AI answers is in scope, the report should keep the questions, platforms, dates and repeat conditions. A few favorable screenshots are no substitute.
After making the correction locally, we checked the resources and navigation again. Those checks confirmed that the fix worked in the local version. They did not establish that it had been published, that search engines had recrawled it or that inquiries had increased.
Explore our GEO audit if you need this kind of review. For a first check of your own, start with the crawler-access guide.
Updated: · published: · Reviewed by the geo-rank.ai editorial team · Our methodology