GEO vs SEO

The short version

SEO gets you into an ordered list of links. GEO gets you named inside a written answer. Most of the underlying work is shared — both depend on pages existing, being readable, and being corroborated elsewhere — but the metric, the volatility and the pages that decide the outcome are different.

Treating one as a rename of the other leads to measuring the wrong thing, which is the expensive mistake here. Nobody loses money by doing both; people lose money by reporting one as if it were the other.

Side by side

What carries over

More than the “SEO is dead” posts suggest. Assistants read the open web. Pages that were worth ranking are the same pages worth citing, and the technical groundwork is identical.

Crawlability and indexing. A page an engine cannot fetch is not a candidate for anything. Blocked paths, slow responses, JavaScript-only content and broken canonicals hurt both surfaces equally.

Page structure. Clear headings, one idea per section, lists and tables rather than walls of prose. What made a page scannable for a human makes it parseable for a model.

Factual accuracy and specificity. Concrete numbers, named services, real prices. Vague pages rank badly and get cited rarely, for related reasons.

Topical depth. A site that covers a subject properly is a better retrieval candidate than one with a thin page per keyword. This was already true for search; it is more true now.

Internal linking. Still how a crawler finds things, still how importance is communicated.

Reputation off-site. The thing SEO called link building overlaps heavily with what matters here, except that the currency is being described accurately rather than being linked.

If you have been doing competent SEO, you are not starting from zero. You are starting from a position you cannot yet see.

What does not carry over

Keyword-shaped thinking

Buyers type keywords into search boxes and sentences into assistants. Where do I go for an emergency dental appointment in Odesa occupies a different phrasing space from dentist odesa, and a question set built from a keyword export misses it.

This is not a small difference in wording. Question phrasing carries intent — urgency, price sensitivity, comparison — and intent is exactly what determines which businesses an assistant considers worth naming.

Rank tracking

There is no rank. A tool that reports “position 4” for an assistant answer is reporting an invention, and a report built on that number is worse than no report because it looks legitimate.

The replacement is a share across repeats, reported per assistant. It is a less tidy number and it is the only honest one.

The assumption that the homepage is the whole job

This is the expensive one. In one local market we measured, a single assistant answer rested on eleven pages on average — your own pages plus directories, review sites, local roundups and trade listings.

For SEO, your site is the asset and the lever. For GEO it is also the reference copy: the facts a buyer asks about — prices, services, areas, availability — have to be stated somewhere you control before anything else can repeat them correctly, and the pages that get quoted are the ones answering a specific question rather than the homepage. What is added is the second half of the job: making the sources read alongside your pages agree with them. That half is slower, because it depends on somebody else accepting a correction.

Averaging

SEO taught people to report an average position. Do that across assistants and you produce a number describing none of them.

When four assistants were asked the same questions in one market, they named 63 businesses between them. Only four were named by all four. Forty-six appeared in a single assistant only. In the same measurement, one business ranged from 100% of answers on one assistant to 50% on another.

An average would have reported something in the middle, which was true of nothing and actionable for no one.

The click as proof of work

SEO could always point at sessions. An assistant may name you without linking, and the buyer arrives days later by typing your name — recorded as direct traffic with no explanation. If sessions are your only success measure, this work will look like a failure while succeeding.

Where the two disagree in practice

Three situations where the same page performs differently on each surface.

A page that ranks well and is never cited. Usually because the answer is buried. A page that builds to its conclusion serves a reader who scrolls and fails an engine extracting two sentences.

A page that is cited constantly and ranks poorly. Usually a short, direct answer on a low-authority domain. Retrieval systems care less about domain strength than a ranking algorithm does.

A business that ranks nowhere and is named anyway. This is the one that surprises people most. If the directories and roundups an assistant reads describe a business well, it can be named without its own site figuring at all.

Should you move budget from SEO to GEO?

For most businesses this is not a real choice: they are the same budget line doing two jobs, and the overlap in the work is large.

The practical sequencing:

1. Keep the SEO that works. Especially the technical groundwork and the pages that bring qualified traffic.
2. Measure where you stand in assistant answers. You cannot plan without this, and it is cheap.
3. Spend the marginal effort on the question types where you are absent, which usually means sources rather than pages.
4. Report the two separately. Never add a share of answers to a ranking.

What we would avoid is replacing a working SEO programme on the strength of a category that is two years old and cannot promise placement. Anyone recommending that is selling a rebrand.

A worked example

From a measurement on one local market, business anonymised.

Search position: top three for the obvious commercial keyword, stable for over a year. By SEO reporting, the account was healthy.

Assistant answers: named in 38% of general recommendation answers and 0% of price answers, across four assistants.

What the citations showed: for price questions, the same four sources repeatedly — two directories, a review platform, a regional roundup. The business appeared on one of them, with an out-of-date service list and no prices.

What was done: nothing to the homepage. Listings corrected and completed on all four sources, using the service names and price ranges from the business’s own pages. One page rewritten so the price question is answered in the first two sentences.

What moved: general recommendation shifted four points, which we reported as no change, because that is what four points is. Price answers went from zero to roughly a third.

What we did not claim: that the listings caused it. We reported what was changed, when, and what moved after — and that the movement was concentrated in the question group the work targeted rather than spread evenly. Weak evidence, honestly labelled, is worth more than a confident story.

Note what this example is not: an SEO failure. The rankings were fine throughout. The two surfaces were simply answering different questions about the same business.

Frequently asked

Is SEO dead? No, and the people saying so were mostly saying something similar about voice search in 2018. Assistants read the open web; pages remain the raw material.

Do I need a separate agency for GEO? You need someone who measures it. If your current supplier can show you a share of answers per assistant with the raw answers behind it, they are already doing the work.

Can my rank tracker do this? No. See above: there is no rank.

How long before GEO results show? Weeks to months, and it is gated by other people’s queues rather than by crawl schedules.

Which matters more? Depends entirely on whether your buyers ask assistants. The cheapest way to find out is to ask your own buying questions and see whether the assistants name anybody at all.

How to report both without confusing anyone

The most common failure after starting this work is not technical — it is a report that mixes the two surfaces and leaves the reader unable to act.

Keep them in separate sections. Search performance in one, assistant answers in the other. Never in the same table.

Never add them. A position and a share are different kinds of number. Averaging or combining them produces a figure that means nothing.

Use different words. “Rank” belongs to search. “Share of answers” belongs to assistants. Using “visibility” for both is how the confusion starts.

Label the surface on every figure. Not “38%” but “38% of general recommendation answers on ChatGPT”. The extra words are the whole point.

Report drift as drift. Under about five points on the assistant side is not progress, and saying so in the report is what makes the other numbers believable.

A short glossary

Rank — a position in an ordered list of results. Exists in search, does not exist in assistant answers.

Share of answers — the percentage of answers naming a business across repeats. The assistant-side metric.

Citation — a source shown with a generated answer.

Drift — variation caused by the system rather than by work.

Questions people actually ask

If assistants read the open web, is GEO just link building again? Closer than most vendors admit, with one difference: the currency is being described accurately rather than being linked. A directory that lists your services correctly helps even with a nofollow link, because the text is what gets read.

My rankings are fine and assistants ignore me. What is wrong? Almost always that your off-site description is thin while your own site is strong. Search rewards your pages; assistants read what others write about you.

Assistants name me and my rankings are poor. Should I care? Depends who your buyers are. If they ask assistants, you are already winning the surface that matters to them.

Do I need two suppliers? No, and two suppliers is how you end up with two reports that disagree. You need one who measures both and reports them separately.

Which should a new business do first? Ordinary SEO, because it builds the pages everything else reads. Then measure the assistant side to find out which sources decide your category.

Where to start

Measure once, so the two surfaces can be compared rather than conflated.

• The free AI visibility check runs a short measurement and emails you what the assistants said, with no call.
• The method explains how the question set is built and why every figure is per assistant.
• AI search optimization is the full guide, and how to improve AI visibility is the step-by-step version of the work.

A note on suppliers

Two things worth checking in any proposal that mentions both disciplines.

Does it use the word “rank” about assistant answers? If so, the supplier has not understood the surface, and the reports will contain numbers that cannot exist.

Does it report per assistant? A single blended “AI visibility score” is the clearest signal that the measurement is not built on anything you can act on. Ask to see the four numbers behind it, and the raw answers behind those.

One thing to do tomorrow

Open your last SEO report and your buying questions side by side. Ask each question once in an assistant. If the businesses named are not the ones ranking, you have just learned the most useful thing in this article, and it took five minutes rather than a quarter’s budget.