Archives: Knowledge Base

Can we get incorrect information about our business removed from AI answers?

Publish an unambiguous, well-structured correction on your own domain first. Correct third-party listings carrying the wrong information; corroboration across sources is what…
What is the difference between AEO and GEO?

In practice the technical work overlaps substantially — both depend on extractable, corroborated, retrievable content. The distinction matters most for measurement: extraction…
How many AI engines should we track?

Engine relevance is category-dependent — check where your buyers actually are before paying for coverage. Each engine cites differently, so results are…
How do you build a prompt set?

Twenty to forty prompts is sufficient for a single-service business; multi-vertical companies need more. Include prompts where you would expect to lose,…
Should we block AI crawlers?

Blocking is often accidental — security plugins and CDN rules block unknown user agents by default. Check your robots.txt and firewall logs…
Does page speed affect AI visibility?

The critical question is whether your content exists in the delivered HTML without JavaScript execution. Test by disabling JavaScript and reloading. Anything…
What makes content extractable by AI systems?

Front-load the answer. Content that builds to a conclusion after 600 words of preamble is rarely extracted. One question per section, with…
Why do AI answers differ between users asking the same question?

This is why single spot-checks are worthless as measurement — one query is one sample of a distribution. Reliable measurement requires repeated…
How often should AI visibility be measured?

Measure the same prompt set each time. Changing the set mid-programme invalidates every comparison you have. Take two baseline readings a week…
What is an AI Presence Score?

Built from three inputs: citation frequency, position within the answer, and prompt coverage across the tracked set. Scores move when models update,…