Getting ChatGPT to recommend your school: a GEO and AEO guide for education

Getting ChatGPT to recommend your school: a GEO and AEO guide for education

Heavy use of AI tools to search for information rose from 5% to 21% between 2023 and 2025 according to SparkToro. Traffic referred by AI assistants tripled in six months to around 6% of the total according to SEMrush. And in 2026 the share of searches that end in a generated answer without a visit to any website already exceeds the share that produce a click.

Translated to a school's situation: a family moving to your city and asking "which bilingual schools are in this area and which have a good reputation" increasingly asks an assistant, not a search engine. And the assistant answers with three or four names. This article is about being among those names.

What GEO and AEO are, without the marketing

Traditional SEO aims to appear in a list of links. The two sibling disciplines change the objective:

• GEO (generative engine optimisation): the answer a model builds should come from your content and carry your name.
• AEO (answer engine optimisation): structuring content so it becomes the chosen answer.

In practice they overlap so much that separating them is not worth the effort. What matters is the shift in objective: it is no longer only about earning a click, but about earning a mention. And a mention inside a generated answer has a property a link does not: it arrives with the implicit authority of having been selected.

First: entity consistency

A language model does not hold a record of your school. It reconstructs who you are from agreement between sources. When those sources contradict each other, the model settles on the most repeated version, which is not always the right one.

That is why the highest-return work is not on your website but off it. Check that the official name, address, phone number, stages, status — state, state-funded or private — and languages match exactly in five places:

• Your website.
• Your Google listing.
• Education directories.
• Your regional education authority's portals.
• Any comparison platform where you appear.

This is also the remedy for the most frequent problem: an assistant giving outdated information. It almost always traces back to an old listing in a directory nobody remembers creating.

Structured data: what models read well

Structured data markup is not a ranking trick, it is a way of saying things unambiguously. Three types matter for a school:

• Educational organisation: declares what you are, where you are, which stages you offer and how to reach you.
• FAQs: turns each question and answer into a citable unit. It gives the most return per unit of effort, because it matches the way people actually ask.
• Events: open days and admission deadlines, exactly the kind of concrete information an assistant looks for when answering.

Citable content: answer-first format

A model cites fragments that answer something. It does not cite introductory paragraphs. That changes how you write:

• Citable: a section starting "the admission window in this region usually opens in March and runs for three weeks".
• Not citable: one starting "choosing a school is one of the most important decisions".

The useful structure puts the answer in the first or second sentence and develops afterwards, rather than building towards a final conclusion.

The local questions nobody else answers

One type of content is especially valuable for a school: the kind that answers local, concrete questions nobody else answers:

• How admission scoring works in your region.
• What documentation is needed.
• How school transport operates in your area.
• What meal subsidies exist.

These are questions families genuinely ask and about which there is very little specific content.

Reviews and third-party mentions

When an assistant assesses reputation, it does not invent it: it takes it from what other sources say. Google reviews, mentions in local press, appearances in education portals, participation in sector events.

The practical consequence is that asking satisfied families for reviews — something almost no school does systematically — now has a double return: it influences the families who read them and the models that process them. And the best way to get them is to ask just after a specific positive experience, not in June with a generic email.

How to measure whether it works

There is not yet a standard tool equivalent to Search Console for this, so do it by hand and consistently. Prepare a list of ten or fifteen realistic questions from a family in your area, put them to two or three different assistants once a month and note three things:

• Whether they mention you.
• Whether the information is correct.
• Which source they cite when they cite one.

Three months of records already show a pattern, and above all they surface the serious errors, which is the immediate benefit of the exercise.

Case study (Spain)

A state-funded school started running this monthly check and discovered two things in the first round. It appeared in answers about schools in its neighbourhood, but was described as a private school with no mention of its public funding, which completely changed the price expectation. And its sixth-form provision, added three years earlier, appeared in no answer at all.

The source of both errors sat outside its website: two education directories with 2019 listings. Correcting them took two emails and three weeks of waiting. In the following month's check the description was already right.

The second change was content: they published four articles answering concrete questions about the admission process in their region, with the answer in the first sentence of each section and FAQ markup. Two months later they began appearing as a cited source in answers about admissions, not just about their own school.

Conclusion

The channel through which families discover a school is shifting, and it is shifting faster than most schools have registered. The good news is that competition is still low: almost nobody in education is working on this, and the highest-return actions — fixing inconsistent listings, marking up FAQs, answering concrete local questions — need no budget, they need consistency.

At Edena we apply exactly this to our own content, with structured FAQs across more than four hundred articles in five languages. Book a demo and, alongside the platform, we will tell you what has worked for us and what has not.