What 6,187 SEO Optimisation Records Reveal About AI Search Readiness
Digital Womble research and analysis, 19 November 2026
The headline: AI search visibility is not a file-upload exercise. It starts with websites that search systems can discover, interpret and trust. Our own audit records reveal how much basic work remains, while independent AEO research shows why a Google ranking alone cannot guarantee visibility across AI engines.
Our original data: 57 completed audit runs
We reviewed stored records in Digital Womble's On-Page SEO Tool on 19 November 2026. The dataset contains 57 completed audit runs across 48 distinct website hostnames, with 6,219 page discoveries and 6,187 recorded page optimisations. These are run-level totals, not 6,187 unique pages or evidence that 6,187 changes have been published to live sites. A repeat scan may count the same page again.
Of the 57 runs, 28 contain page-level metadata scores, providing 1,730 scored page instances covering 1,465 distinct page URLs. In that scored subset, 319 of 1,730 (18.4%) scored below 50 and 682 (39.4%) scored below 70 on the tool's metadata-length compliance measure. The average score for these instances was 69/100. These are not comprehensive SEO quality scores: the historic measure reflects title and description length heuristics. The sample is drawn from our tool's users and must not be interpreted as representative of all UK business websites.
| Metric | Recorded result | Interpretation |
|---|---|---|
| Completed audit runs | 57 | Completed tool records |
| Distinct website hostnames | 48 | Not necessarily distinct client businesses |
| Page discoveries | 6,219 | Counts across runs |
| Recorded optimisations | 6,187 | Proposed/generated optimisation output, not confirmed deployment |
| Runs with page-level scores | 28 | Subset suitable for metadata analysis |
| Scored page instances | 1,730 | Includes repeat URLs |
| Distinct URLs within scored subset | 1,465 | URL-level deduplication count |
| Instances below 70 on metadata-length compliance | 682 (39.4%) | Not an overall SEO failure rate |
What this actually tells us
There is a substantial volume of basic optimisation work in the websites reaching our tool. But it would be misleading to claim that metadata length alone determines ranking or AI citation success. A good title can still target the wrong intent. A technically sound page can still fail to answer a buyer's question. And a well-written page may still be missing the external evidence that makes a business credible.
That is why we are separating metadata compliance, overall SEO quality, commercial opportunity and observed AI visibility in future reporting rather than presenting one convenient but misleading score.
What the independent AEO research adds
Paul Andre de Vera's State of AEO 2027 research at aeo.so, with data collected in November 2026, analysed 2,126 AI answers. It reported 3–17% pairwise source-domain overlap between engines in its studied sample. The study also reported that YouTube featured in 66% of sampled Google AI Overview answers and official sources accounted for 59% of ChatGPT citations in its dataset. These findings are specific to the study's AEO-related queries and should not be extrapolated to every industry.
The commercial implication is not that every business must produce more YouTube videos. It is that different answer engines can rely on different evidence, and being prominent in one discovery channel does not ensure inclusion in another.
Source: Paul Andre de Vera, State of AEO 2027, https://aeo.so/research/state-of-aeo. Figures above are attributed to the original publisher, not presented as Digital Womble findings.
Are llms.txt and llms-full.txt worth adding?
They can be useful optional machine-readable guides, particularly for systems that choose to consume them. But neither file guarantees citations, inclusion or ranking in ChatGPT, Gemini, Perplexity or Google AI Overviews. Do not confuse an accessible summary file with a proven visibility signal. Keep XML sitemaps, canonical URLs, robots directives and genuinely useful HTML pages correct first.
Five practical checks before paying for another AI visibility tool
- Can crawlers access the pages that matter? Check robots directives, indexability, canonicals and XML sitemaps.
- Does each important page answer a real buyer question? Compare title, heading, content and search intent, rather than relying on word counts.
- Can visitors and crawlers find your strongest pages? Put relevant contextual internal links naturally inside article paragraphs, including an early link to the most useful service or supporting resource.
- Is there evidence behind the claims? Use named authors, genuine case studies, verifiable figures and relevant third-party citations.
- Are you measuring actual outcomes? Separate recommended fixes from published fixes, organic impressions from clicks, and AI mentions from citations or enquiries.
For an actionable assessment, start with Digital Womble's on-page SEO optimisation service, then review our AI visibility approach. If your issue is commercial rather than purely technical, see our managed search growth service.
Methodology and limitations
Digital Womble figures are a descriptive snapshot of application records accessed on 19 November 2026. Runs marked as tests were excluded where flagged. Hostnames were normalised by removing the www prefix. Page-instance percentages use 1,730 stored scores, not the 1,465 deduplicated URLs; therefore repeat pages may influence the percentages. No ranking uplift, citation uplift, revenue gain or live deployment is inferred from optimisation output. Independent aeo.so figures are separately attributed and concern a different sample and research question.
Bottom line: AI visibility is an outcome to measure, not a technical box to tick. Improve the website foundations, publish useful evidence, earn credible references and then test whether search engines actually surface your business.