The four fifths that are the same job
Before the interesting part, the boring part, because a lot of money is currently being spent on the boring part twice.
An answer engine finds your page the same way a search engine does. It crawls the open web, or it queries a search index, and it needs the page to be reachable, readable and about something.
Every fundamental of technical SEO carries straight across: server-rendered HTML, sane site structure, working internal links, fast pages, no gating, a sitemap, correct canonicals, and a robots.txt that does not accidentally exclude the crawlers you want.
The content fundamentals carry too. Clear writing about a specific thing, organised under headings that match the questions people ask, with the claims stated rather than implied. Good SEO content and good retrieval content look nearly identical, which is why teams with a decent content programme tend to appear in answers without having tried.
If your SEO is weak, so is your GEO. They stand on one foundation, and the extra work goes on top of it rather than around it.
Five places they genuinely diverge
| SEO | GEO | |
|---|---|---|
| Unit | The page ranks | A passage gets retrieved. A strong page with no standalone sentences contributes nothing. |
| Competition | Position. First takes most of the clicks, eighth takes almost none. | Inclusion. Being the fourth of six sources still puts your name in the sentence. |
| Authority signal | Links pointing at you | Independent sources describing you the same way |
| Query shape | Keywords, short, typed | Questions, long, conversational, often comparative |
| Measurement | Search Console, impressions, position, click-through | Referral traffic and manual sampling. No console exists. |
The authority row is the one with strategic weight. SEO taught a generation of marketers that authority is a thing you accumulate by acquiring links to your domain.
Retrieval does not work that way. What decides how a model describes you is how many independent pages describe you, and whether they agree.
Twenty write-ups all landing on the same sentence about what you do is worth more than a link from a site with a high domain rating, because the model is synthesising descriptions rather than counting endorsements.
That distinction is why the two disciplines eventually stop being the same job. The mechanics are here.
The payoff curve favours challengers
In search, the reward for position is brutally steep. The first result takes the overwhelming majority of the attention and the eighth takes a rounding error, which is why a new company can do everything right for a year and see almost nothing. You are either above the incumbent or you are invisible.
An answer names three to eight companies in a flat list. Fourth place in that list is a mention in the same sentence as the market leader, read by a buyer who is assembling a shortlist. The gap between first and fourth is small in a way it has never been in search.
For a company that cannot outrank an incumbent with eight years of backlinks, this is the most useful structural fact in the whole subject. The bar moved from beating them to being mentioned alongside them, and the requirement for that is enough independent sources treating you as part of the category.
The click that does not arrive
The honest cost of all this: a good deal of the value never shows up as a session.
A buyer asks which tools do the thing, reads the answer, notes three names, and closes the tab. You were named. Nothing happened in your analytics.
Later that week they type one of those three names directly, and by then the visit looks like direct traffic with no story attached to it.
Two consequences worth being clear-eyed about.
- The attribution case is weaker than SEO's. Anyone reporting a confident figure for answer-engine influence is reporting a model, not a measurement. Referral traffic from the engines is real and undercounts the effect badly.
- The intent case is stronger. The sessions that do arrive come from someone who read a summary and is clicking to verify one specific thing, usually pricing or whether you handle their case. They convert at rates that make the small volumes worth having.
Treat it the way a sensible team treats being named in a category roundup. It moves consideration, it resists precise measurement, and the correct response to both facts is to keep doing it while refusing to invent a number for the board.
Where doing one hurts the other
Mostly they reinforce. Four places they pull against each other, and all four are common.
- The comprehensive guide. SEO rewards long pages that cover a topic exhaustively. Retrieval rewards a compact sentence that answers one question. A 4,000-word guide that ranks well can still contribute nothing to an answer if its conclusions are distributed through narrative. Fix by front-loading each section with its claim, which costs the ranking nothing.
- Keyword page proliferation. Spinning up thin variants for long-tail terms is an old SEO tactic with modest upside. It has direct downside here, because a scatter of near-duplicate pages describing the company slightly differently degrades the consistency the model needs.
- Gating. Content marketing teams gate the comparison guide to capture emails. Comparison content is close to the highest-value thing a retriever can find about you, and a gate makes it invisible. This trade used to be defensible and is not anymore.
- Blocking crawlers. The instinct to block AI crawlers for content-protection reasons removes you from the answers entirely. There is an argument for blocking training crawlers. There is none for blocking the search crawlers that fetch a page in order to cite it.
Whether to buy it as a service
The market for GEO retainers arrived faster than the discipline did, and most of what is being sold is a rebadged SEO audit with a section about ChatGPT at the front.
The question to ask a prospective vendor is narrow: what will you do that our existing SEO scope does not already cover? Anything about crawlability, schema, page structure, internal links or content quality is the SEO scope.
If the answer is a rank-tracking dashboard, note that there are no ranks to track, and that a tool sampling ten prompts a week is doing manually what your own team could do in an hour with better questions.
The part that is genuinely additional is the authority row from the table above, and it is not optimisation work at all. Getting twenty independent sources to describe your company in the same words means giving twenty people a reason to write about you in the same week, which is a launch.
That is the half we do, and it happens to be the half that also produces the pipeline, which is the correct order to think about it in.
Do the site work yourself or with whoever already does your SEO. It is finite, it is cheap, and it is a weekend of effort for most teams. Then spend the budget on the reason for anyone to write anything.
The diligence questions are the same ones you would ask about any agency, with one addition: ask them to show you a company whose description in the answers changed after they worked on it.
Questions we get asked
Is GEO a separate discipline from SEO?
Roughly four fifths of the work overlaps. Both need crawlable pages, clear content, a consistent entity and other sites describing you.
The genuine differences are the unit of retrieval, the shape of the payoff, the missing click and the absence of any measurement tool. Those belong inside the programme you already run rather than in a second one.
Should we hire a GEO agency?
Ask what they will do that your SEO scope does not already cover. The site work is the same work.
The half that genuinely differs is getting other people to describe you consistently across crawlable pages, which is distribution rather than optimisation. A retainer that produces a rank-tracking dashboard for a channel with no ranks is worth declining.
Does GEO bring traffic?
Less than SEO does, and at higher intent. Many buyers read the answer and never click, so the value is being named in a consideration set the buyer assembles in thirty seconds. The sessions that do arrive convert unusually well because the visitor has already read a summary and is coming to verify one thing.