Case study · SEO, AEO and GEO
kuraib.site audit case study: what a self-audit found, and what changed
I audited my own site, published the result, then went back and checked every finding against the live pages.
By Kuraib AliPublished Updated
In July 2026 I audited my own site, kuraib.site, the way I audit a client's: technical SEO, on-page content, readiness for AI answers (AEO and GEO), and usability. I published the result with a score. This is the same case study, rewritten on 11 November 2026 after I went back to the live site and checked each finding again.
Two things changed in the rewrite. I removed the scores, because I had assigned them myself and published no method for them. And I removed every outside statistic I could not trace to its source. What is left is what a reader can verify: three gaps, and the state of each one today.
What the July 2026 audit covered
The audit read four areas, in the order the rest of the work depends on:
- Technical foundation. robots.txt, the sitemap, canonical URLs, and whether crawlers can read the pages.
- On-page content. One H1 per page, a heading order that makes sense, and the search phrase in the title and the first paragraph.
- AEO and GEO readiness. Whether the pages state who is behind them, give facts that can be checked, and are open to the crawlers that AI answers rely on.
- Usability. Whether a visitor can read the page and find the next step.
The same four areas are what my technical SEO and AI search optimisation work covers for clients.
The three gaps it found
| Gap in July 2026 | Why it mattered | Status on 11 November 2026 |
|---|---|---|
| No structured data anywhere on the site | Markup tells a search engine in plain terms who the business is and what a page is about | Closed |
| No Wikidata entry for the business owner | An entry is one more place that states who a person is | Still open |
| No published method or original material to cite | A page that only repeats others gives nobody a reason to reference it | Partly closed |
Gap 1: structured data, now closed
In July the site had no structured data. On 11 November 2026 the live pages carry it:
- The home page has Person, WebSite and WebPage markup.
- The services page has Service and BreadcrumbList markup.
- The about page has Person and BreadcrumbList markup.
- The work page has CollectionPage markup.
- The audit offer page has Service and FAQPage markup.
The original plan was to add FAQ markup to five pages. I would not write that plan today. Which markup still earns a rich result has narrowed, and my guide to schema types that still get rich results explains where FAQ markup stands. On this site I use none.
Gap 2: Wikidata, still open
I searched Wikidata for "Kuraib Ali" on 11 November 2026 and found no entry. The July plan called this a quick win. It is not: Wikidata has its own rules about who qualifies for an entry, and creating one for yourself is the wrong way round. I have left it open and I am not counting it as a task.
Gap 3: something worth citing, partly closed
The July audit said the site had no original method on record. Since then the visibility stack, the five-layer order I work in, has been written up in full and now lives on this site as the visibility stack. The site's work page lists eight case studies. A published ranking case study is still missing, and I say so on my service pages.
What changed that the audit did not predict
The crawler policy was reversed for training bots. The July audit praised the robots.txt file for allowing GPTBot. The live file now does the opposite for crawlers that collect training data: GPTBot, ClaudeBot, CCBot and Google-Extended are disallowed. The crawlers that fetch pages for AI search and for a user's own request, including OAI-SearchBot, ChatGPT-User and Claude-SearchBot, are still allowed. This site, kuraibseo.site, allows both kinds. Neither choice is the correct one for every business, and the guide on robots.txt and AI crawlers sets out what each bot does so the choice can be made on purpose.
The llms.txt file stayed, with lower expectations. The audit rated the site's llms.txt highly. The file is still there and it is accurate. But Google has said its Search does not use llms.txt, so I no longer describe it as a strength. The evidence is in does llms.txt work for AI search.
What I would not claim again
- A score out of ten. A number I gave my own site proves nothing. A list of gaps with a dated status does more.
- Borrowed statistics. The first version quoted figures about AI search from studies I could not find again when I looked for the source. They are gone.
- Comparisons with other sites. The first version said how other agencies' sites tend to score. I had no data for that.
What this shows about an audit
An audit is useful when someone else can check it. The parts of the July audit that survived were the plain findings: this is missing, here is where, here is what fixes it. The parts that did not survive were the ones that sounded most impressive. When I audit a client's site, the report is the first kind: a short list of faults, where each one is, and the order to fix them in. The free SEO audit follows that format.
Frequently asked questions
Is a self-audit score evidence of anything?
Not on its own. The first version of this page gave kuraib.site a score out of ten that I assigned myself, with no published scoring method. I have removed it. What a reader can check is the list of gaps and whether each one is closed on the live site, so that is what this page now shows.
What did the first version of this case study get wrong?
It led with scores nobody else could reproduce, it quoted industry statistics I could not trace back to a source when I re-checked them, and it compared the site with other agencies' sites without data. All three are gone. The findings themselves held up: structured data was missing, and it has since been added.
Does an llms.txt file help a site appear in Google's AI answers?
Google says its Search does not use llms.txt, and that there are no extra files or markup required to appear in AI Overviews or AI Mode. kuraib.site has one, and I treat it as a convenience for tools that read it, not as a ranking lever.