GEO and AEO: a practical audit for AI search visibility

GEO usually refers to improving visibility in generative answers. AEO covers answer-oriented search and assistants more broadly. These are overlapping industry terms, not two technical standards. The practical goal is to make relevant information discoverable and correctly understood.

• What AEO services include

• How to measure AEO results

Choose a page and a customer question

Start with a real question such as ‘Can I move my customer database without losing its history?’ Check whether the page explains file formats, supported fields, limitations and validation. Markup cannot supply missing information.

Check technical access

• Open the public URL without login. Check errors, redirects and incorrect canonicals.
• Inspect indexing and display restrictions in the relevant search engine’s tools.
• Check the main text in the fetched response and, for JavaScript-heavy sites, in the rendered page.
• Review robots.txt, CDN and server blocks. Allow verified crawlers as documented rather than disabling protection.

OAI-SearchBot and PerplexityBot serve search. Allowing GPTBot for training is a separate choice. Follow each provider’s documentation instead of one blanket rule for all AI bots.

• OpenAI crawler controls

• Perplexity crawlers and verified IP ranges

Make the answer complete

Lead with the answer, then explain conditions, examples and exceptions. Let the question determine length. There is no need to squeeze a complex explanation into 40–60 words for an assumed algorithm.

Example: importing customer data

‘Fast migration without data loss’ gives a reader little to assess. Explain accepted formats, which fields transfer, unsupported data, duplicate handling and who checks the result. Use your product’s actual limits. Add an anonymised sample file and validation steps, then link to relevant pricing and terms.

Check facts, responsibility and translations

• Identify the author or responsible team and cite technical sources.
• Give case-study figures a period, sample size and measurement method. Separate observation from explanation.
• Change the update date when the content materially changes, not merely to look recent.
• Review each translation for meaning, currency and regional service conditions. hreflang tells systems that support it which pages are equivalent language or regional versions. It does not determine which language version an AI service will cite.

Markup, FAQs and llms.txt

Structured data gives systems a machine-readable description of what is on the page, such as an article’s author and dates. It must match the visible content. Adding more schema types does not make the information more useful, and Google does not require special GEO markup or llms.txt for its generative search features.

• Google’s AI search optimization guidance

Use FAQs when readers have outstanding questions. Google’s changelog says FAQ rich results stopped appearing in May 2026. A useful Q&A section still helps readers, but it does not promise a special search display or an AI citation.

• Google’s FAQ rich-result update

• When to implement llms.txt

Accept the work and measure changes

1. Save the original question, page state, service answer and cited URLs.
2. After publication, check access, new links, mobile layout and agreement between markup and text.
3. Repeat the same questions under comparable settings on different days. Keep answers without citations too.
4. Track visibility, visits and enquiries separately, noting other changes that happened during the same period.

The audit should identify missing information and access problems; implementing its recommendations should resolve them. A later increase in mentions is an observation. A single before-and-after test cannot establish what caused it or predict sales.

• A method for measuring ChatGPT answers

• Choosing between SEO and GEO tasks