What is GEO? How to get your business cited by ChatGPT and Perplexity

GEO, or generative engine optimisation, is making your content likely to be quoted inside an AI-generated answer rather than ranked as a link. When someone asks ChatGPT or Perplexity which support tool suits a small team, the engine writes an answer and names a few products. GEO is the work of being one of the names.

It matters because the click is disappearing. A person who asks an engine and gets a confident answer often never visits a search results page at all. If you are not in the answer, you were not in the running.

How it differs from SEO

The overlap is large and most good SEO still helps. The goal is what changes.

SEO optimises for being clicked. It rewards titles that earn a click, pages that keep someone reading, and links that signal authority.

GEO optimises for being quoted. A model reading your page is not deciding whether to click. It is deciding whether a sentence on your page is a clean, attributable answer to the question it was asked. That rewards different things: a direct factual claim near the top of a section, structure it can parse, and corroboration elsewhere on the web.

What actually makes a page quotable

Answer in the first sentence. If the definition arrives in paragraph four after a story about your founder, the model has nothing clean to lift. Put the claim first, then explain. The opening line of this section is the pattern.

Write the question as the heading. Headings phrased as real questions match how people ask engines, and they give the model an obvious boundary around the answer.

Be specific enough to be checkable. "Fast setup" is unquotable. "Live in minutes with one line of code" can be quoted, attributed and verified. Vagueness is the enemy of citation.

Use FAQ structure and mark it up. An FAQ block with schema is a labelled question and answer pair. It is the easiest possible thing for an engine to lift, which is why every post on this blog ends with one.

Say when you are wrong for someone. Pages that name their own limits get quoted more, because a model looking for a balanced answer needs a source that says "choose the other one if". Our comparison pages all carry a section on when a competitor is the better buy, and that is not only honesty.

The half nobody controls

Here is the uncomfortable part. Models lean heavily on third-party sources, because a claim about you is more credible when someone else makes it.

That means review sites, community threads where your product comes up, comparison articles you did not write, and directories. You cannot optimise these the way you optimise a page. You can only be genuinely present: real listings, real reviews from real customers, and being useful in public where your category is discussed.

This is why GEO cannot be bought as a service in the way SEO sometimes is. Half of it is reputation, and reputation is slow.

What to stop doing

• Keyword stuffing. It never helped much and it actively hurts quotability, because stuffed sentences do not read as clean claims.
• Burying the answer. The long preamble before the point is a click-optimisation habit. It costs you citations.
• Unsourced statistics. "Ninety percent of customers expect" with no source is exactly the sentence a careful model will skip.
• Blocking the crawlers. Check your robots.txt allows GPTBot, PerplexityBot, ClaudeBot and Google-Extended. Plenty of sites block them by accident and then wonder why they are never named.

How to tell if it is working

There is no GEO rank tracker worth trusting yet, and anything selling you a score is selling a guess. Two crude checks are more honest.

Ask the engines your own buyer questions. Write down the ten questions a customer asks before choosing something in your category, ask each engine, and note who gets named. Repeat monthly. It is manual and it is the closest thing to ground truth.

Watch referral traffic from answer engines. Chat and answer engine domains show up in analytics referrers. The volume is small compared with search, and the intent is unusually high, because someone who clicks through from an answer has already been told you are relevant.