Specialist SEO
Generative Engine Optimisation Services That Get Your Brand Recommended
Ask a chat assistant to recommend a supplier in your field and it will name a few businesses with confidence. I work on why it names those and not you: how clearly your business is described, where you are mentioned, and whether the pages it reads give it a reason to put you on the list.
- years in SEO
- 6+
- client websites worked on
- 60+
- person doing the work
- 1
Generative engine optimisation for answers that name names
Generative engine optimisation is the work of getting your brand named, recommended and cited when people ask ChatGPT, Perplexity, Gemini or Copilot which provider, product or approach to choose. Those answers read like advice from a well-informed friend, and they usually mention a short list of businesses. If you are not on that list, the buyer may never reach your site, however well it ranks in Google.
A chat assistant has two ways of knowing about you. One is what the model absorbed in training from the open web: articles, reviews, directories, forum threads and your own pages, as they stood at the time. The other is what it retrieves live when the product searches the web before replying. GEO works on both. The first is slow and indirect; the second responds faster to changes on your site and on the pages that get retrieved.
This page covers recommendations and citations inside chat assistants. Winning the quoted reply box on a Google results page is a different, more structural job under answer engine optimisation. The broader programme that combines both, with crawl access and measurement built in, is AI search optimisation.
Key takeaways
- Chat assistants build answers from two places: what the model learned in training and what it retrieves live. GEO works on both, through your own site and the third-party sources the assistants read.
- Being recommended depends heavily on being described consistently. If your site, profiles and press coverage all say slightly different things about what you do, an assistant has less confidence naming you.
- Shortlist and comparison prompts are often answered from list articles, review platforms and directories, so earned presence on those sources matters as much as your own pages.
- Mentions and citations are measured by running a fixed set of real prompts repeatedly, because answers shift with session, wording and platform.
- There is no form to submit and no fee that buys a place in an organic chat answer. The work is gradual, cumulative and mostly about being the obvious choice in the sources that get read.
What's included in my generative engine optimisation work
Six parts. The recommendation audit sets the balance: some brands are left out because nobody describes them clearly, others because they are missing from the sources assistants read.
A recommendation audit of who chat assistants name in your category
I run the prompts buyers would use to find a provider - 'best X for Y', 'alternatives to Z', 'X vs Y', 'who does X near me' - across ChatGPT, Perplexity, Gemini and Copilot. For each answer I log which brands are named, in what order, how each is described, and which sources are linked or appear to be drawn on.
- Brands named per prompt and platform
- How each competitor is described
- Sources linked or referenced
- Prompts where you are missing or misdescribed
Brand entity work so models describe what you do, and for whom, correctly
Assistants summarise whatever they find about you. I write one canonical description - name, category, services, locations, audiences, what sets you apart - and bring your homepage, about page, Organization schema, Business Profile, LinkedIn and industry directories into line with it. Where an assistant gets a fact wrong, I trace the source it most likely came from and fix that source.
- A canonical brand description agreed with you
- Site, schema and profiles brought into line
- Wrong or outdated facts traced back to a source
- sameAs links between your official profiles
Mentions on the sources generative answers keep drawing on in your sector
From the audit I list the domains that keep appearing behind answers in your category: trade publications, review platforms, directories, comparison articles, forums and community threads. Then I work out where a mention is realistic - a listing, a review profile, an expert quote, a correction to an outdated roundup - and pursue those, without paying for placements dressed up as editorial.
- A ranked list of influential sources
- Listings and review profiles claimed or completed
- Expert commentary pitched where it fits
- Outdated roundups flagged for correction requests
Shortlist visibility for 'best', 'top' and 'alternatives' style prompts
Prompts that ask for a shortlist reward brands that are easy to compare. I make sure your site has honest comparison pages where they are justified, pricing or pricing guidance, stated use cases, and a plain account of who you are not right for. Those pages hand an assistant the facts it needs to place you, and they reassure buyers who arrive from the answer.
- Comparison and alternatives pages where justified
- Pricing guidance and use cases stated plainly
- 'Best for' positioning written in your own words
- Specific claims a model can repeat accurately
Retrieval access so live-search assistants can actually read your key pages
Several assistants search the web at the moment they answer. I confirm that the crawlers they use for search are allowed in robots.txt, that important pages render without depending on heavy scripts, and that the site is indexed in Bing as well as Google, since some products rely on Bing. Training-crawler choices are made with you separately and on purpose.
- Search crawler access confirmed per assistant
- Rendering checked on the pages that matter
- Bing and Google indexing compared
- Training crawler policy agreed separately
Citation and mention tracking across a fixed set of buyer prompts
I rerun the same prompt set on a schedule and report your share of mentions, citations of your pages, how you are described and which sources gained or lost ground. For a few dozen prompts I do this by hand; at larger scale it runs as a browser pipeline with a screenshot per answer under AI search visibility tracking.
- Mentions and citations per prompt over time
- Description accuracy noted per platform
- Referral visits from assistants in analytics
- Screenshots kept as a record of each run
Experience
Generative Engine Optimisation from someone who has done it many times
You work with me from the first call to the last report. No account managers, no hand-offs to a junior team.
Book a free call- years in SEO
- 6+
- client websites worked on
- 60+
- documented case studies
- 35
Which chat assistants I cover and what shapes their shortlists
Each assistant assembles a recommendation differently, and none publishes the full method. This is the working picture I plan from.
| Assistant | How it tends to build a recommendation | GEO priority |
|---|---|---|
| ChatGPT | Model knowledge, plus a web search when the question calls for current information, with sources linked | Search crawler access, consistent brand facts, presence in the roundups and reviews it retrieves |
| Perplexity | Searches the web for almost every question and lists its sources prominently | Pages that state facts plainly, and mentions on the sites it cites for your category |
| Gemini | Google's models grounded in Google Search and Google's knowledge of entities | Healthy Google indexing, an accurate Business Profile, entity facts that match across sources |
| Microsoft Copilot | Bing search results combined with the model's own knowledge | Bing indexing, Bing Webmaster Tools, and the publications Bing ranks for your topics |
| Claude | Model knowledge, plus web search where it is available and switched on | The same consistent facts and third-party presence that help with the others |
None of these companies explains how brands are picked for a recommendation. The table reflects their own documentation where it exists and what I see when running prompts, and I re-check it at the start of every engagement.
How a GEO engagement runs
The audit and the entity fixes fill the first few weeks. Source work carries on after that, because mentions build up slowly.
- 1
Agree the prompt set
Twenty to fifty prompts that real buyers would type, from category questions to direct comparisons, fixed so later runs compare like with like.
- 2
Run the baseline
Each prompt runs several times on each assistant, because one answer is an anecdote. Brands, order, descriptions and sources all get logged.
- 3
Fix the description at source
Your own site, schema and profiles are corrected first, since they are fully in your control and the fastest thing to change.
- 4
Map the influential sources
Domains that recur behind answers are ranked by how often they appear and how realistic a mention on each one is.
- 5
Earn mentions and corrections
Listings, reviews, expert quotes and correction requests, month by month, aimed at the sources the baseline surfaced.
- 6
Rerun and report
The same prompts on a schedule, with changes in mentions, citations and descriptions shown next to referral visits.
Free review
Get a free Generative Engine Optimisation review
Send me your site. Before the call I look at your rankings and pages, then tell you what I would fix first and what it is likely worth.
Two ways in
Training data moves slowly, live retrieval moves faster
What a model learned in training changes only when a new version is trained, so improving how the web describes you today pays off later and unevenly. Live retrieval is different: when an assistant searches before replying, a corrected profile, a new review page or a fresh roundup mention can show up within weeks. I plan the work so the fast changes land first and the slow ones keep compounding behind them.
Generative engine optimisation claims to be wary of
A new acronym brings a new set of sales pitches. These are the ones I hear most, and what I would do in their place.
Common pitches
- Guaranteed placement in ChatGPT answers within a set number of days
- Paid slots in 'best of' lists that exist for crawlers, not readers
- Mass-produced articles stuffed with your brand name
- A visibility score built from a handful of prompts run once
- Hidden text or instructions aimed at steering AI crawlers
What holds up
- No promises about answers that no outside party controls
- Earned mentions on sources real readers use
- A small number of pages that state your facts clearly
- Large prompt sets rerun over time, with their limits stated
- Visible, honest content written for the buyer first
What I measure for GEO and how to read each number
No platform reports this for you in full. These are the measures I put in front of clients, with the caveat that goes with each.
| Measure | How it is collected | How to read it |
|---|---|---|
| Share of mentions | Fixed prompts rerun across assistants on a schedule | The direction over several runs, never the figure from a single day |
| Citations of your pages | Links shown alongside answers, logged per prompt | Shows which of your pages assistants treat as evidence |
| Description accuracy | Each assistant's summary of your business, checked against the canonical description | A wrong fact points to a source that needs correcting |
| Source share | Domains linked behind answers in your category | Tells you where the next mentions will matter most |
| Assistant referrals | Analytics sessions from assistant domains, plus 'how did you hear about us' answers | Undercounts, because many answers send no click at all |
Who gains most from GEO
People lean on chat assistants most for research and shortlisting. These are the situations where that shows up in enquiries.
Considered B2B purchases
Buyers ask assistants for vendor shortlists and comparisons long before they fill in a form. Missing from the shortlist means missing from the deal.
Software and online services
Category, alternatives and integration prompts are answered heavily from review sites and roundups, where presence can be earned.
Professional services
People ask which kind of specialist they need and who is well regarded. Clear positioning and reputation sources decide who gets named.
Newer or rebranded businesses
Models trained before your launch or rename may describe an old version of you. Live sources are the practical way to correct that.
How GEO connects to your other SEO work
Most of the source work overlaps with reputation and authority building. Campaigns aimed at links rather than mentions sit under link building, and wider brand signals such as reviews and press under off-page SEO. If assistants cannot retrieve your pages in the first place, the fix starts with technical SEO.
For examples, see how a language services company had pages hidden from search opened up, and how a private-markets investor club built content around investor questions. To follow your brand across hundreds of prompts rather than a few dozen, the measurement runs as AI search visibility tracking.
Growth from real projects
Learn more about generative engine optimisation
Often needed alongside this
Generative Engine Optimisation Services FAQs
What is generative engine optimisation?
It is the practice of making your brand more likely to be named, recommended and cited when people ask generative AI assistants such as ChatGPT, Perplexity, Gemini or Copilot for advice. It combines clear, consistent information about your business, presence on the third-party sources those systems read, and pages that state your facts plainly enough to be repeated accurately.
Is GEO just SEO with a new name?
There is a large overlap: assistants that search the web rely on pages being crawlable and indexed, and on many of the same authority signals. The difference is in what you are trying to win. SEO aims for a ranked link; GEO aims for a mention inside a written answer, which depends more on how other sites describe you and less on any single page of yours.
Can you get us recommended by ChatGPT?
I cannot promise it, and nobody outside OpenAI can. Answers vary by user, wording and session, and the way brands are chosen is not published. What I can do is find out why competitors are named instead of you, fix how your business is described, earn presence on the sources answers draw on, and measure whether your mentions grow.
How long does generative engine optimisation take to show results?
Changes that affect live retrieval, such as a corrected profile or a new mention on a frequently cited site, can appear within weeks on assistants that search the web. Changes in what models know from training depend on new model versions and are slower and less predictable. I expect a few months before the prompt set shows a trend you can rely on.
Do we need to block or allow AI crawlers for GEO?
For assistants that search the web, blocking their search crawler can keep your pages out of their answers. Training crawlers are a separate choice: some businesses allow them, some do not, and some vendors let you control the two independently. I go through each one with you, using that vendor's documentation as it stands today, and set the policy deliberately.
What if an assistant says something wrong about us?
First I find where the claim probably comes from: an outdated directory listing, an old press article, a competitor's comparison page or your own old pages. Correcting that source, and publishing a clear, current statement on your site, is the most reliable route. Feedback buttons on answers are worth using too, though they are not a fix on their own.
Get a free generative engine optimisation snapshot for your category
Tell me the three prompts you would most like to be recommended for. I will run them across the main assistants, show you who gets named and which sources sit behind those answers, and outline the first steps to get your business onto that shortlist.