The method, not the pitch. Complete without hiring anyone.
Generative engine optimization — the practical guide
How to get your brand named inside AI-generated answers: what actually determines citation, the three levers you can pull, how to measure whether any of it worked, and the mistakes that waste the first quarter.
On this page
- Why this is a different job to SEO
- The three levers
- How to actually run it
- What not to bother with
- Questions, answered
Why this is a different job to SEO
In classic search the engine returns a list and the user chooses. Your job is to occupy a high position in that list. In generative search the assistant reads a set of sources, synthesises an answer, and names a few of them. The user often never sees a list at all, so a position is not what you are competing for — inclusion in the synthesis is.
That changes what a good page looks like. A page that ranks well is often long, comprehensive and written to cover a topic exhaustively. A page that gets cited is one where a specific, quotable claim can be lifted cleanly out of it and attributed. Those two things overlap, but they are not the same, and optimising only for the first leaves citations on the table.
It also changes where the work happens. A meaningful share of what assistants cite is not your own site at all — it is review sites, industry roundups, forum threads and comparison articles written by other people. In classic SEO those are competitors. In GEO they are frequently the surface you need to appear on.
The three levers
Everything worth doing in GEO falls into one of these.
- 01
Extractability — structure your own content to be quotable
A direct definition in the first hundred words. Specific claims with named sources rather than vague superlatives. Comparison tables. Clear headings that match how a question is actually asked. The test is simple: could a paragraph be lifted out of this page and still make complete sense on its own?
- 02
Presence — appear on the sources assistants already trust
Find which domains get cited for your category’s questions, then work out how to legitimately appear on them: review platforms, industry roundups, credible forum discussion, comparison content written by third parties. This is digital PR work, and in concentrated categories it outweighs everything you do on your own site.
- 03
Access — make sure the crawlers can read you
AI crawlers are separate from Googlebot and are frequently blocked by default in a robots.txt nobody has read in three years. Check that the crawlers you want are allowed, that your content is not locked behind rendering they cannot perform, and that your schema markup is valid.
How it runs
How to actually run it
In order. Most of the wasted effort in GEO comes from starting at step four.
- 1
Write and freeze a prompt set
Twenty to fifty buying-intent questions, phrased the way a customer would type them. Freeze the list — measurement is only comparable over time if the questions stay identical, and an improved prompt set resets your history.
- 2
Take a baseline
Run the set across the assistants that matter to you and record which name you, which name competitors, and which name nobody. Without this you cannot tell later whether you improved or the model simply changed.
- 3
Map the citation landscape
For every prompt, record which domains were cited. A pattern emerges quickly, and it is usually concentrated: a handful of domains take most of the citations. That list is your target list.
- 4
Fix access before writing anything
Check crawler permissions and rendering. Publishing content that the assistants you care about cannot read is the most common wasted quarter in this discipline, and it takes an afternoon to rule out.
- 5
Restructure your highest-intent pages
Start with the pages closest to a buying decision, not the blog. Add the definition, the sourced claims, the comparison table. This is the cheapest work with the fastest feedback.
- 6
Go after the cited domains
Get listed, reviewed, included in the roundups, present in the discussions. Slower and harder than editing your own pages, and in concentrated categories it is the only thing that moves the number.
- 7
Re-measure on the frozen set
Same prompts, same method, recorded with the date. Change in citation share on a fixed set is the only honest measure of whether the work is doing anything.
What not to bother with
The advice circulating that we would skip.
- 01
Writing "for the AI" in a special voice
There is no separate register that models prefer. Clear, specific, well-sourced writing is what gets extracted, which is the same thing that makes a page good for humans. Anyone selling a special AI writing style is selling a style.
- 02
Chasing every assistant equally
Your buyers use one or two. Measure which, and concentrate. Spreading the same effort across every assistant produces a thin result everywhere.
- 03
Publishing volume for its own sake
Citation is concentrated on sources that say something specific. Forty generic pages give a model forty reasons to cite someone else. One genuinely authoritative page outperforms them.
- 04
Treating it as a replacement for SEO
The overlap between cited pages and well-ranking pages is substantial. Abandoning classic search to chase citations removes the foundation the citations are partly built on.
Run the method, or have us run it
The guide above is complete — you can execute it without us. If you would rather not, thirty minutes gets you a prompt set, a baseline and a prioritised list of what to change first.
FAQ
Questions buyers ask us first.
Is GEO just SEO with a new name?
There is real overlap — both reward clear, credible, technically accessible content — but the target differs. SEO competes for a position in a list; GEO competes to be quoted inside an answer. The practical differences show up in how you structure a page and in how much of the work happens on domains you do not own.
Do I need to do this if my SEO is already strong?
Strong SEO gives you a substantial head start, since ranking pages are cited disproportionately often. It is not sufficient on its own: pages that rank first get passed over, and pages outside the top ten get cited. The gap between those two facts is what GEO works on.
How do I know if it is working?
Citation share on a frozen prompt set, measured on a cadence and dated. Not traffic — assistants often do not send a click, so judging this by referral traffic will tell you it failed even when it is working.
Can I do this myself?
The first two levers, largely yes. Restructuring your own pages for extraction and fixing crawler access are both within reach of a competent marketer, and this guide is written to be enough. The third lever — earning presence on the domains that get cited — is slower and is where most people bring in help.
What if assistants cite nobody in my category?
Then the answer is generic guidance and there is no vendor to displace, which is a genuine opportunity and a slower project. The work becomes creating the sourced, specific material a model would need in order to name anyone at all — and the first credible source in a category tends to keep that position.
What happens next
- Answer a few quick questions about your business. About two minutes.
- Pick a time that suits you for a free 30-minute call with a founder.
- We study your site before the call, so you leave with a plan, not a pitch.
Free, no obligation. Your answers save as you go, so you can stop and pick up where you left off.