Every few years the industry finds a new acronym and sells packages against it before anyone agrees what it means. GEO is the current one. Underneath the marketing there is a real change, and for local businesses it is narrower than the pitch decks suggest.
What GEO means once you strip the hype
Generative engine optimisation is the practice of making a business likely to be named in an answer that a model writes, rather than in a list of links a search engine returns. Someone asks an assistant for a good physiotherapist in Sheffield who can see them this week, and gets three names in a paragraph. Whether a business appears in that paragraph is the whole of GEO.
What it is not: a new algorithm to reverse engineer. There is no index to submit to, no console, no set of directives. Models answering local questions read the open web plus whatever structured place data they can reach, then summarise. The lever is what they find.
Citation, not ranking, is the unit of success
This is the part that trips up people coming from classic SEO. There is no position one. An answer either mentions the business or it does not, and if it does, the mention comes with a characterisation: "well reviewed for emergency work", "the cheaper of the two", "closed on Mondays". The characterisation is as important as the mention and it is assembled from source material you may not have looked at in years.
Some consequences that follow directly:
- Being second does not exist - a two-name answer includes you or it does not. Marginal improvements in relevance matter less than being unambiguously the right kind of business for the question.
- Adjectives come from reviews - if reviewers describe a business as fast and expensive, that is the description that gets generated. Nothing on the website overrides it.
- Contradictions cost you - a model that finds two different phone numbers has a confidence problem and often resolves it by naming a different business instead.
- Nothing is trackable in the old way - answers vary by phrasing, by user, by session. Sampling a set of realistic questions monthly is the honest approach. Anyone selling a GEO rank tracker is selling a fiction.
Where generative answers source local facts
For a local question, the material being drawn on is fairly consistent. Roughly in order of how much weight it tends to carry:
| Source | What it supplies |
|---|---|
| The Google Business Profile itself | Name, address, hours, category, services, attributes, review content |
| The business website | What it does, where, for whom, at what price, in your own words |
| Review platforms | Sentiment and specific descriptions from customers |
| Directories and industry bodies | Corroboration of the basic facts, plus credentials and memberships |
| Forums and community threads | Unfiltered opinion, often quoted almost verbatim in answers |
| Local press and listings | Recency, and evidence the business is real and active |
The pattern is that generated answers favour facts that appear consistently in several independent places. A claim that exists only on the business's own site is a claim; the same claim echoed across a professional register, a directory and three reviews is a fact.
Consistency across the open web
This is where the actual work lives, and it is unglamorous enough that most competitors will not do it.
Start with the facts that must never disagree: legal name, trading name, address format, phone number, opening hours, service area. Then check the places they appear. Old directory entries from a previous address, a Companies House record with a different trading name, a Facebook page nobody has logged into since 2019 - each of these is a contradicting source, and contradictions are exactly what reduce the confidence that gets a business named.
Beyond the basics, three things make a business legible to something summarising it:
- Plain declarative statements on the site - "We are a family-run electrical contractor covering Leeds and Bradford, working on domestic rewires and consumer unit upgrades." One sentence that a machine can lift. Most homepages open with something that says nothing.
- Service pages that name the service - one page per real service, using the words customers use, with location stated in the body text rather than only in the footer.
- LocalBusiness structured data - not because a model requires it, but because it removes ambiguity about which entity the page describes. Cheap to add, hard to get wrong.
Review content deserves specific attention here. Generic five-star reviews contribute nothing to a description. Reviews that mention the service performed, the area, and something concrete about the experience are the raw material for how a business gets characterised. That does not mean scripting reviews, which breaks Google's policies and reads badly to humans too. It means asking at the point where the customer has something specific to say.
Doing this across a handful of locations is a spreadsheet job. Across fifty it is not, and the value of a tool like Ampli5 Pulse here is boring rather than clever: one place to correct profile fields across every location, and reviews in a single queue so replies actually happen.
What GEO does not replace
A local business still gets most of its customers from people who are physically nearby and searching in the normal way. Proximity has not stopped mattering, the map pack has not disappeared, and a profile that ranks badly for the obvious query is not rescued by being mentioned in a chatbot answer.
The honest framing for a client is that GEO is a second front, not a replacement. Nearly everything that makes a business citable in a generated answer also makes it stronger in conventional local search: accurate structured data, service pages that say what the business does, corroborating references across the web, reviews with substance. If someone proposes a GEO programme that requires abandoning that work, the proposal is the problem.