Answer Engine Optimisation (AEO) is the work of making public content easier for search and AI systems to discover, understand and use in answers. It builds on SEO foundations and useful content. It cannot guarantee indexing, citations or the accuracy of an answer generated by another system.
Start with the Question Your Page Answers
A service page should explain what is offered, who it helps, what a customer receives, its limitations and how to proceed. An explanatory article should answer its central question early, support material claims and provide examples. Clear headings and relevant internal links help readers navigate the subject.
For example, a grant eligibility page needs the actual criteria, dates, exceptions and responsible organisation. A short summary is useful only if the supporting detail remains accessible and current. There is no universal answer length that makes a page eligible for AI search.
Google's optimisation guidance keeps established search practices central. Bing's AI Performance guidance also recommends clear structure, evidence and consistent entity descriptions. Neither establishes a formula for guaranteed citations.
Check Discovery Before Adding Markup
- Confirm important pages return the intended HTML and can be reached through crawlable links.
- Check canonical URLs, indexing directives and sitemap destinations agree.
- Check that main content is accessible and useful, with sources and dates where relevant.
- Inspect account-level search settings and actual crawler logs when access is available.
A successful request using a bot user-agent string is only a simulated access check. Infrastructure can treat genuine crawler requests differently. OpenAI's crawler documentation distinguishes OAI-SearchBot for search discovery from GPTBot for potential model training; one policy should not be used as a proxy for the other.
Describe Content Accurately with Structured Data
Use schema that matches the page: an Article for an article, or a Service entity for a service description. Keep organisation and author identities consistent and link them to real public information. Dates and URLs should agree with the page's metadata.
Schema describes claims; it does not verify facts or stop hallucinations. Adding FAQPage, HowTo or Speakable everywhere is not a general citation technique. Check current feature support and Google's structured data policies before implementing a type.
Semantic HTML and a logical heading hierarchy also help accessibility and navigation. They should be implemented for those concrete benefits, without presenting perfect markup as a proven AI ranking requirement.
Worked Example: TrueCMS's Published Feed
Our 5 September 2026 audit of TrueCMS checked 39 sitemap URLs and the resources advertised by the site. The pages returned HTML with canonical links and parseable JSON-LD, but the machine-readable interface had separate problems:
| Observed technical finding | Correction and acceptance check |
| --- | --- |
| The page-data feed omitted the AEO service and actual page content. | Generate records from the content source and verify that the AEO record agrees with its visible page. |
| The feed advertised /content/site-config.json, which returned 404. | Remove the broken promise or replace it with an intended public resource; verify every advertised link. |
| The plugin manifest advertised resources on an old domain that redirected to the homepage. | Use the correct resource URLs and check the returned content and response type, not just a successful status. |
This is a first-party example of diagnosing an interface defect. The audit did not measure indexing, citations or conversions, and these findings do not demonstrate ranking uplift. A reliable feed serves its consumers; publishing a custom feed does not itself establish a Google search benefit.
Google explicitly says llms.txt does not improve its visibility or rankings. Prioritise working public pages and existing interfaces before optional agent-specific exports.
Measure Outcomes Separately from Technical Checks
Technical acceptance checks establish that a correction works. Search performance requires a separate baseline and follow-up observations:
- Where available, Google's Generative AI performance report provides impressions for supported AI search features. It is not a complete citation or conversion report.
- Bing AI Performance provides citation activity and grounding-query information. Citations do not establish clicks or sales.
- Analytics can track referral traffic and qualified enquiries. Compare equivalent reporting periods and document other changes that may affect the result.
- Manual prompt checks need a recorded engine, prompt, date, locale and cited URL. A single answer cannot establish general visibility or causation.
When account access is unavailable, say what remains unverified. Do not replace absent performance evidence with a successful schema validation result.
Put the Findings into an Implementation Plan
Prioritise inaccessible or contradictory information, then improve usefulness and evidence, correct metadata, and establish repeatable measurement. Our AEO audit and implementation service sets out the scope, deliverables, access requirements and limits of that work.
