There are two bad positions in this argument. One is that generative engine optimisation is an entirely new discipline that makes SEO obsolete — usually argued by people selling GEO...
There are two bad positions in this argument. One is that generative engine optimisation is an entirely new discipline that makes SEO obsolete — usually argued by people selling GEO retainers. The other is that it is just SEO with a new label — usually argued by people who have not measured anything. Both are wrong, and the useful answer is boring: the foundation is shared, and four things genuinely differ.
If a page cannot be crawled, rendered and understood, no engine of any generation will use it. Indexability, site speed, clean URL structure, structured data, sane internal linking, and being a credible source on a topic all carry over unchanged. Roughly, the majority of the work behind AI visibility is technical and editorial work that an experienced SEO team already knows how to do. Any provider claiming their GEO methodology is wholly novel is charging you a premium for a fundamentals audit.
| Classic search | Generative answers | |
|---|---|---|
| What you win | A ranked position on a results page | A sentence inside an answer |
| How many win | Ten blue links, plus features | Often three to five brands, sometimes one |
| Click outcome | Position drives the click | Many answers resolve without any click |
| Measurable by | Rank tracking, impressions, clicks | Mention Rate, Citation Rate, Share of Model |
The practical consequence: position 8 is worth something in classic search and worth nothing in an answer. Generative surfaces are far more winner-concentrated, which makes a mid-table position a much weaker outcome than the equivalent ranking.
This is the finding that surprises clients most. When we ran a 37-prompt category set and looked at where the model’s vendor names came from, the pattern was consistent: names came from third-party sources — listings, directories, press, community discussion — not from the vendors’ own marketing pages. Your own site controls whether you can be cited once you are already known. It has far less influence on whether you are known at all.
The work this implies is closer to digital PR and entity consolidation than to on-page optimisation: consistent naming across every profile that mentions you, presence in the sources that cover your category, and a clear, machine-readable statement of what you do and where you operate.
In classic SEO, a large glossary of definitional articles is a defensible traffic strategy. In generative search it is close to worthless. In our own run, only 6 of 37 prompts produced any vendor name, and every single one was a buying-stage prompt — comparisons, selection criteria, pricing questions, “who should I hire” questions. Top-of-funnel and mid-funnel prompts named nobody at all.
If your content plan is ninety percent “what is X” articles, it is optimised for a channel that is shrinking and irrelevant to the channel that is growing.
A ranking is a fact you can look up. A generative answer is a sample from a distribution: the same prompt can produce different answers, and the model can change underneath you without notice. This is not a reason to skip measurement — it is a reason to measure like a statistician rather than like a rank tracker. Fixed prompt set, repeated runs, aggregate rates rather than single observations, and a competitor board so you can tell “we moved” apart from “the model moved”.
Fix the technical foundation, because it serves both. Rebalance the content plan towards buying-stage material. Invest in third-party presence, which is the lever with the clearest observed effect. And measure both channels separately, because a single blended score will hide whichever one is failing.
No. The majority of the underlying work is the same, and classic search still sends the volume. It is a rebalancing of effort, not a replacement.
Only if your current team cannot show you a measurement method. The skills overlap heavily; the discipline that is genuinely missing in most teams is measurement, not optimisation.
It costs almost nothing, so add it if you like. Just do not confuse it with a strategy — we have not seen evidence that it materially moves mention or citation rates.
Third-party coverage takes months to accumulate and models take time to reflect it. Expect two to three measurement cycles — six to nine months — before movement is clearly attributable.
Want your own numbers instead of ours?
Send us your domain. We run the baseline on your category prompts and send back the raw answers, not just a score.
Đội ngũ chuyên gia Vidco Group sẵn sàng đồng hành cùng bạn
Bước 1 / 4
Chúng tôi sẽ liên hệ trong vòng 2 giờ làm việc.