The GEO guide
GEO (generative engine optimization) is the work of making AI answer engines more likely to name and cite your product when people ask relevant questions. The practical loop is: measure what AI says today, find the gaps, publish pages written for those questions, and retest.
What GEO is
GEO is about the answer, not the list of links. When someone asks ChatGPT, Perplexity, Claude or Google’s AI Overview for a recommendation, the model writes an answer and may name products and cite sources. GEO is the practice of influencing whether you are one of them.
It is not a replacement for SEO. In our own data the pages that rank well are largely the pages that get cited, so the two reinforce each other. What GEO adds is a new thing to measure: what AI says about you.
Why it can be nonlinear
Paid acquisition is linear: spend stops, traffic stops. An AI answer that names you works differently. Every person who asks that question, today and next month, sees the same recommendation. Content that earns that position keeps working without more spend.
This is a mechanism, not a guarantee. AI answers change, and nobody controls them. The point is that the effort compounds when it works.
The loop: measure, gap, publish, retest
Run the same loop every month so you can see what moved.
- Measure. Build a bank of questions in three groups: brand, user intent, competitor comparison. Send them to the major models and record who is named and cited.
- Find the gaps. Where you’re absent, where a competitor is named, where you’re mentioned but not cited.
- Publish. One page per gap, answer-first, with sources.
- Retest. Same bank next month. Keep what moved the numbers.
Mentioned is not cited
AI can name your brand without linking to you (a mention), or link to your page without naming the brand in the answer (a citation). They need different fixes. A mention-only pattern usually means AI knows you from elsewhere but your own pages don’t answer the question. A cite-only pattern means your pages are useful but your brand isn’t associated with the category.
Track both, separately.
What to write
Write for the questions people actually type, in their words. Reddit, forums and customer calls are the dictionary. Then format for extraction:
- Answer first: a direct 1–2 sentence answer at the top of the page
- Tables and lists for comparisons and steps
- Numbers with a source and a date
- A short FAQ at the end
- Clear author and updated date
- Consistent product name and description everywhere
Where to be present
Your own site is the base. But AI also leans on third-party sources, so being discussed on communities, reviewed on comparison pages and mentioned by other sites matters. If AI names you without citing you, this is usually where the gap is.
Measure without fooling yourself
AI answers vary from run to run. Ask each question more than once and track rates, not single answers. Keep the question bank fixed so months are comparable, and add new questions as a separate group. Don’t treat one good answer as a win.
Common mistakes
Writing generic “what is X” pages with no answer-first paragraph. Chasing mentions for brand-only questions nobody asks. Changing the question bank every month. Publishing a lot and never retesting.
Our own result
We run this loop on JobWizard with a bank of about 50 questions. In September 2026 that gave 46% of answers naming the brand and 88% citing our pages, with no ad spend. Read the case.
FAQ
What is the difference between GEO and AEO?
They are used almost interchangeably. AEO (answer engine optimization) emphasizes answer engines; GEO emphasizes generative models. The work is the same.
How long until AI starts naming me?
It varies by market and how much content exists for your questions. Measure monthly so you can see movement, and don’t promise yourself a date.
Do I still need SEO?
Yes. Ranking pages tend to be the pages that get cited, and search remains a major source of traffic.