GEO

Being the source an AI assistant names.

GEO is the work of becoming a source that generative systems retrieve, summarise and cite when they answer a question in your category. It is a real discipline with a peer reviewed origin, and it is surrounded by more nonsense than any other part of marketing right now.

2024The year the term was defined in a paper
Up to 40%Visibility gain measured in that paper
0Vendors who can guarantee a citation
Assistants this work is measured against: ChatGPT, Google Gemini, Perplexity, Siri and Amazon Alexa.
A contour terrain model built from stacked layers of pale card, with a single slender brass pin standing at its highest point.
In short

What is generative engine optimization?

Generative engine optimization is optimising content to be used and cited by systems that generate an answer rather than return a list of links, such as Google's AI Overviews and AI Mode, ChatGPT, Perplexity and Copilot. The term comes from a 2024 paper by Aggarwal and colleagues, presented at KDD, which measured methods improving visibility in generative engine responses by up to 40%.

It is not a separate technical discipline bolted onto SEO. Google states plainly that there are no additional requirements to appear in AI Overviews or AI Mode, no special optimisations, and no AI specific files or markup needed. What changes is the objective: being summarised accurately and named, rather than ranked and clicked.

Nobody can guarantee a citation in an AI answer. Outputs vary between sessions for the same question and no vendor offers a placement mechanism. What can be done is make the inputs better and measure what actually comes back.

How an AI assistant builds its answerA chat panel where someone asks which business to use nearby, and the assistant replies with a short shortlist. Three arrows run back from the answer to the sources it read: the website, the map profile and the reviews.AI ASSISTANTWHO SHOULD I USE NEAR ME?BASED ON WHAT I CAN READ ABOUT THEMYOUR BUSINESSHOURS, SERVICES AND REVIEWS ALL MATCHANOTHER OPTIONFEWER DETAILS PUBLISHEDANOTHER OPTIONDETAILS DISAGREE BETWEEN SITESWHAT IT READSYOUR WEBSITESERVICES · AREAS · HOURSYOUR MAP PROFILENAME · ADDRESS · PHONEYOUR REVIEWSWHAT CUSTOMERS WROTEIT CAN ONLY REPEAT WHAT YOU HAVE ALREADY PUBLISHED.
How an AI assistant builds its answer
Plain language

What actually changed

For twenty five years, search returned a list of links and the goal was to be near the top of it. Increasingly, search returns a written answer with a handful of sources attached. The goal has moved from being clicked to being used.

This is not a prediction. It is measurable. SparkToro's analysis of United States Google searches using Similarweb clickstream data found 68.01% of searches ended without a click between January and April 2026, up from 60.45% in 2024. Two thirds of searches now resolve without anyone visiting a website.

That does not mean websites stopped mattering. It means the website's job changed. It is now the source material a system reads in order to say something about you, as well as a destination someone might reach.

If your site is thin, contradicts your own listings, or cannot be read without running JavaScript, it is poor source material. That is the practical core of GEO, and it looks a great deal like good SEO done carefully.

Measured, not predicted

The click really is disappearing

SparkToro analysed United States Google searches using Similarweb clickstream panel data. This is behavioural measurement of real browsing rather than a survey or a model.

68.01%of US Google searches ended without a click, January to April 2026
60.45%the same measure in 2024

SourceZero-click search study, SparkToro with Similarweb clickstream data, 2026

The trend line is the point. If your entire marketing measurement rests on sessions, you are measuring a shrinking share of the attention your brand actually receives.

Where the term came from, and what the research found

GEO was defined in a 2024 academic paper, GEO: Generative Engine Optimization, by Aggarwal, Murahari, Rajpurohit, Kalyan, Narasimhan and Deshpande, presented at the KDD conference. This matters because almost nothing written about GEO online cites it, and a great deal of what is written contradicts it.

The paper built a benchmark of real queries, ran them through generative engines, and tested which modifications to source content changed how visible that content became in the generated answer. The headline finding is in the panel below, with a link to the paper itself.

That finding lines up with our own experience and with Google's public position, and it is worth stating plainly: the way to be used as a source is to be a good source. Cite things. Be specific. Say who is responsible for the claim.

Primary source

What the research actually reported

From the paper that introduced the term, presented at the KDD conference. This is the source almost nothing written about GEO cites, and it is the only peer reviewed measurement available.

Up to 40%improvement in visibility in generative engine responses using GEO methods
2024the year the term was defined and the methods were tested

SourceAggarwal et al., GEO: Generative Engine Optimization, arXiv and KDD, 2024

The methods that performed well were adding citations to authoritative sources, including relevant statistics, and quoting credible people. Methods amounting to keyword manipulation performed poorly.

Who is already doing it

This is present tense, not a forecast

The CMO Survey is run by Duke University's Fuqua School with Deloitte and the American Marketing Association, and surveyed 308 United States marketing leaders in January 2026.

4 in 10companies already using generative engine optimization
308US marketing leaders surveyed, 97% vice president level or above

SourceThe CMO Survey, Duke Fuqua with Deloitte and the AMA, 2026, 308 US marketing leaders

Which is the honest reason we treat this as a live part of a search program rather than as an emerging trend to discuss next year.

Unglamorous, mostly

What the work actually consists of

If this list looks like careful SEO with different priorities, that is because it is. Google says there is no special optimisation, and the research says the winning methods are about credibility rather than manipulation.

  1. Make the facts about you consistent everywhere

    Your services, hours, locations, credentials and specialisms should say the same thing on your website, your Google Business Profile, your listings and your professional directory entries. A model presented with contradictory sources either hedges or picks one, and it may not pick yours.

  2. Answer questions directly, high on the page

    A question as a heading, a two or three sentence answer immediately beneath it that stands alone if quoted, and the detail underneath that. This is the structural change that makes a page liftable.

  3. Cite things, and be specific

    The GEO paper found that adding citations, statistics and quotations from credible sources improved visibility in generated responses. That is unusually actionable research: a page that names its sources is more likely to be used as one.

    • Named publisher, year and a working link for every number
    • Specifics rather than adjectives, because specifics are what get quoted
    • A named author and a review date, so the page has someone behind it
See the remaining steps: What the work actually consists of3 more stepsHide the remaining steps: What the work actually consists of
  1. Make the page machine readable

    Content present in the HTML rather than only after scripts execute, clean heading structure, structured data that matches what is visible, and no crawl blocking on the pages you want used. Google is explicit that no AI specific files or markup are required, but the ordinary requirements still apply.

  2. Be cited elsewhere

    What other sources say about you carries more weight than what you say about yourself. Professional directories, associations, local press and genuine industry mentions are the levers here, and they are the same levers as digital PR.

  3. Measure what assistants actually say

    Ask the questions a customer would ask, systematically, across the assistants that matter in your category. Record what comes back and which sources get named. There is no dashboard for this, so it is a process rather than a tool.

Notice what is not on this list: no proprietary method, no AI specific file, no guaranteed placement. Those do not exist.

It is everywhere

How to spot GEO nonsense

This is a new enough field that there is no established way for a buyer to tell a real service from an invented one. Here is one.

  • A guaranteed citation in ChatGPT, Perplexity or AI Overviews. No vendor has a placement mechanism, and outputs vary between sessions for the same question.
  • A proprietary AI ranking algorithm or score. Google states there are no additional requirements and no special optimisations for its AI features.
  • A special file you must add to your site for AI. Google states explicitly that you do not need to create new machine readable files, AI text files or markup.
  • A claim about your current AI visibility with no method attached. Ask how it was measured, on which assistants, with which prompts, and how many times.
  • Content published at volume to feed the machines. Google's spam policies name scaled content abuse regardless of how the pages were produced.
See the full checklist: How to spot GEO nonsense3 more itemsHide the full checklist: How to spot GEO nonsense
  • Advice to block AI crawlers as an optimisation. Blocking the crawler that reads your site for an assistant removes you from that assistant's answers. It is a legitimate business decision for some publishers and it is the opposite of GEO.
  • GEO sold as an entirely separate retainer from SEO, when the underlying work is the same work with different emphasis.
  • Any statistic about AI search with no named publisher, year and link behind it. In this field particularly, the invented numbers outnumber the real ones.

The single best test: ask the vendor to show you the primary source for anything they told you. If everything traces to another agency's blog post, you have your answer.

Same work, different goal

SEO and GEO side by side

The overlap is large enough that selling them as separate services is mostly a pricing decision rather than a technical one.

ItemSEOGEO
GoalRank in a list of linksBe retrieved, summarised and named
Success looks likeA position and a clickAn accurate mention, often with no click
Main leverRelevance, depth, linksClarity, consistency, citation by others
Page structureComprehensiveComprehensive, with liftable answers high up
Measurable howRank trackers, Search ConsoleRepeated prompting, recorded over time
Can anyone guarantee itNoNo

A site that is genuinely good at one tends to be reasonably good at the other, which is why we run them as one program rather than two invoices.

Does this work for a local business, and how would you know

It applies, with a different emphasis. When somebody asks an assistant for a recommendation in a place, the answer is assembled largely from structured local data and from what third parties say, rather than from the prose on your website.

That makes your Google Business Profile, your listings and your reviews the primary inputs rather than your blog. A practice whose profile is complete and whose record is consistent everywhere is far more likely to be named than one with a better website and a half filled profile.

Read the full breakdown: Does this work for a local business, and how would you know6 more paragraphsHide the full breakdown: Does this work for a local business, and how would you know

Reviews matter here in a way that is easy to miss, because assistants summarise what people say about you. The substance of your reviews, not just the average, is being read back to prospective customers.

The honest limit: for a local recommendation, proximity and prominence still dominate, exactly as they do in the map pack. GEO does not create a route around geography.

Measuring any of this is harder than measuring search, and we would rather say so than produce a number with no method behind it. There is no equivalent of Search Console for assistants, and anyone showing you a precise AI visibility percentage is showing you their own tool's methodology rather than a platform figure.

What we do is straightforward and repeatable. A fixed set of prompts a real customer might use, run across the assistants that matter in your category, repeated on a schedule because answers vary between sessions, and recorded: were you named, was what it said accurate, and which sources were cited instead.

Alongside that, referral traffic from assistant domains where analytics can see it, which understates the effect considerably because most of the influence produces no click at all. And branded search volume, which remains the most reliable available proxy for whether more people have heard of you.

We report this as an observation log rather than a score, because a score would imply a precision that does not exist.

What we need from you

Accurate facts, and the willingness to publish them. GEO rewards specificity: what you actually do, where, for whom, at what typical cost, with what credentials. A site that will not commit to a specific is poor source material.

Named people. Pages with a named, credentialed author carry more weight with both search engines and generative systems than anonymous ones.

Read the full breakdown: What we need from you3 more paragraphsHide the full breakdown: What we need from you

Consistency discipline. If you change your hours, services or locations, they have to change everywhere, not just on the website.

Realistic expectations about what is controllable. We can improve the inputs and measure the outputs. We cannot promise what a model will say, and we will not pretend otherwise.

We do not publish a price for this piece of work on its own, because the right scope depends on what already exists. What is published is the bundle pricing: 2,400, 3,600 or 4,800 dollars a month depending on which channels are running. You can read the full breakdown on the pricing page, and you will get an exact number in writing before anything starts.

How this connects to the rest

GEO is one piece inside our SEO and GEO program and shares almost all of its foundations with the rest. Technical SEO decides whether your pages can be read at all. Content strategy decides whether there is anything worth citing. Answer engine optimisation is the page level craft of making an answer liftable.

For local businesses, Google Business Profile management and local listings work carry more GEO weight than anything on the website, because that is where the structured facts about you live.

AI search visibility monitoring, under AI implementation, is the measurement side of this work, and the two are usually run together.

Link building and digital PR matter more here than they do for ordinary SEO, because being cited by sources a model trusts is the most direct available influence on what it says.

Find out what assistants currently say about you.

We will run a set of real customer questions through the major assistants before the call and show you exactly what comes back, including who they name instead of you.

Questions

Straight answers.

What is the difference between GEO and SEO?

SEO aims at a position in a list of links, measured by rank and clicks. GEO aims at being used as a source in a generated answer, measured by whether you are named and whether what is said about you is accurate.

The underlying work overlaps heavily. Google states there are no additional requirements or special optimisations for its AI features, which is why we run them as one program.

How do I get my business cited by ChatGPT?

By being a source worth citing: consistent facts across your website and listings, direct answers to real questions, specifics and citations on your pages, and genuine mentions from sources the model reads.

Nobody can guarantee it. Anyone who offers to is describing a mechanism that does not exist.

Is SEO dead because of AI?

No, but the measurement has changed. SparkToro's 2026 analysis found 68.01% of United States Google searches ended without a click, which means click volume understates the value of visibility more every year.

The work that makes you visible is largely the same. What changes is that you should stop judging it purely on sessions.

Does schema markup help with AI search?

It helps machines understand your pages unambiguously, which is useful. Google is explicit that no AI specific markup is required and that structured data must match what is visible on the page.

Treat it as good hygiene rather than as a lever. Anyone selling schema as the route into AI answers is overstating it.

Should I block AI crawlers?

That depends on whether you want to appear in AI answers. Blocking the crawler that reads your site for an assistant removes you from its answers.

For publishers whose business is selling access to content, blocking can be rational. For a local business trying to be recommended, it is self defeating.

How long does GEO take?

The consistency and structure work can change what assistants say within weeks, because these systems retrieve current information rather than working from a fixed snapshot.

Building the third party citations that carry the most weight takes quarters, exactly as it does in ordinary search.

Do AI Overviews reduce my website traffic?

The evidence says yes. Pew Research Center, using a panel of 900 United States adults sharing browsing data across 68,879 Google queries, found 8% of users clicked a traditional search result when an AI summary was present, compared with 15% when it was not.

That is a real cost of this shift and we would rather state it than pretend AI visibility is pure upside.

Next step

Talk to the team

A short call, a look at how the business currently shows up, and a straight answer on what we would do first.