Generative engine optimisation: being in the answer at all.

AEO is about being the extracted answer inside a search results page — an AI Overview or a featured snippet — and is largely driven by how your own pages are written and marked up. GEO is about appearing, and being described correctly, inside answers from assistants like ChatGPT and Perplexity, which draw on both retrieved pages and a wide corpus of third-party sources. GEO therefore involves substantial off-site work that AEO does not.

No, and nobody can. There is no submission process, no ranking factor list and no paid placement, and model behaviour changes without notice. What can be done is measurable: establish what the models currently say, make your own content unambiguous, correct the third-party sources they draw on, decide your crawler access deliberately, and re-measure. That reliably improves presence and accuracy without anyone promising a placement.

It is a genuine commercial decision rather than a technical default. Blocking protects content from being used in training and retrieval, at the cost of visibility in those assistants. Allowing access accepts the former for the latter. What we object to is the common situation where a business wants generative visibility while its robots.txt blocks the crawlers, because nobody made the decision explicitly.

It is an emerging convention for publishing a plain-text, machine-readable summary of a site’s key content at a known path. Adoption by major AI systems is not established, so we treat it as low-cost and speculative: worth adding for a content-heavy site, not worth prioritising over entity consistency and correcting the third-party sources models actually cite today.

Almost always because an outdated third-party source is more explicit than your own site. An old directory entry stating services you dropped years ago is unambiguous, while your homepage may say something aspirational that is hard to parse. The fix is both halves: state plainly what you do on your own site, and correct the external sources that currently contradict it.

With a fixed prompt set, re-run on a schedule across the major assistants, recording whether you appear, how you are described, which sources are cited and who else is recommended. Because the baseline is captured before any work starts, the change is demonstrable — and where a model simply changes behaviour, that shows up too rather than being claimed as a result.

Corrections to your own site and markup can be reflected within weeks. Third-party sources take longer because you are waiting on someone else to publish, and model behaviour changes on its own schedule. Anyone promising a timeline is guessing. We re-measure monthly so you can see movement rather than take it on trust.

In most Midlands B2B niches it is, precisely because nobody else has started. Establishing the correct description of your business now is far cheaper than correcting a wrong one that has propagated. The honest caveat is that we cannot tell you how much traffic it will produce this year.

It is a low-cost, speculative measure. Adoption by major AI systems is not established, so we treat it as worth adding for a content-heavy site and not worth prioritising over entity consistency and correcting the third-party sources models actually cite today.

Then we implement that properly, which is a legitimate and increasingly common choice. Blocking the training and retrieval crawlers costs you visibility in those assistants, and we will be clear about that trade-off rather than letting it happen by accident. What we object to is wanting generative visibility while robots.txt blocks the crawlers.