Contents
I am an AI search and GEO consultant based in London. I help businesses understand and improve how they appear when people ask ChatGPT, Google AI Overviews, AI Mode, Gemini, Perplexity or Copilot for information and recommendations.
Generative engine optimisation (GEO) is the name that has stuck for this work. You will also see it called AEO, AI SEO or LLM optimisation. Whatever the label, the question is the same: when an AI system answers a question your customers ask, can it find you, understand you, trust you enough to cite you, and describe you accurately?
Most of that rests on SEO work that was already worth doing. Some of it is genuinely new, particularly crawler access, third-party reputation and measurement. The useful skill is telling those apart, and not charging you for the parts that have simply been renamed.
Where the experience comes from. As Head of SEO at Frasers Group, I lead long-term SEO and AI search strategy across retail brands including Sports Direct, House of Fraser, Flannels, Studio and GAME. Outside that role, I build the tools I use to test AI visibility and publish what I find, including hands-on testing of ChatGPT Shopping Research when it launched. The tools are free and listed below, so you can judge the thinking before speaking to me.
Who this is for
It is usually worth talking if one or more of these applies:
- Organic clicks have fallen while impressions held up, and AI Overviews now appear on your most valuable searches.
- Competitors are named in ChatGPT or Perplexity answers for your category, and you are not.
- An AI answer describes your business, products or prices incorrectly.
- Leadership wants a GEO or AI search strategy, and nobody can say what it should contain.
- You do not know whether your robots.txt, CDN or firewall is blocking AI crawlers, or whether it should be.
- An agency has proposed GEO work and you want an independent view of what is worth paying for.
- You sell products and want them recommended in AI shopping answers.
It suits businesses of any size. A smaller business usually needs a one-off review; a larger one tends to need ongoing measurement and help getting changes built.
How AI answers find and choose sources
Different AI products work differently, and none of them publishes a full account of how sources are chosen. A few things can be said with reasonable confidence.
Many answers are grounded in live retrieval. When ChatGPT searches the web, or Google builds an AI Overview, the system fetches pages at the time of the question, often drawing on a search index. A page that cannot be crawled or is not indexed is unlikely to be used.
Training data may also play a part. It may influence what a model already associates with your brand, but those influences are difficult to observe and cannot be edited directly.
Answers vary. The same prompt can produce different sources for different users, locations and sessions, which is why a single screenshot proves very little either way.
Google says there is nothing extra to do for its AI features. Its guidance on AI features states there are no additional requirements to appear in AI Overviews or AI Mode beyond its existing Search guidance.
The four layers I assess for AI visibility
I assess AI visibility across four connected layers. A weakness in any one can reduce your chances of being retrieved, cited or described accurately, although no checklist guarantees inclusion.
It can reach the page
Robots.txt rules, CDN bot protection and firewalls can all block AI crawlers, sometimes by deliberate decision and often as a side effect of a security default nobody reviewed. AI companies also run several crawlers with different jobs. OpenAI, for example, documents GPTBot for model training separately from OAI-SearchBot for search, so blocking one is not the same as blocking the other.
It can read the important content
Rendering capabilities differ between crawlers. Googlebot can render JavaScript, but other AI fetchers may not reliably execute it, so content that only appears once scripts run can be missed. I compare the raw HTML with the rendered page to check whether prices, product information, reviews and main copy are consistently available. I have written a step-by-step guide to checking what AI crawlers can see on your site.
It understands who you are and what you offer
Pages that answer the question directly, name products, services, places and people unambiguously, and cover the follow-up questions a conversation produces are easier to quote accurately. This is ordinary good content practice, applied with more discipline about structure and specificity.
Other sources back you up
AI answers often draw on third-party sources: reviews, forums, comparison articles, news coverage and communities such as Reddit. What those sources say can materially influence which brands are mentioned and how they are described, and none of it can be fixed on your own website. That is why I have written about building a genuine presence on Reddit.
What I check, and how
| Area | What I look at | How it is checked |
|---|---|---|
| Crawler access | Whether AI crawlers are blocked, and whether that was a decision or an accident | Robots.txt and CDN bot rules, then verified crawler requests in server logs |
| Rendering | Whether key content is in the HTML or only appears after JavaScript runs | Raw and rendered crawls compared side by side, page source and URL Inspection |
| Indexing | Whether the pages you want cited are indexed at all | Search Console indexing reports and URL Inspection |
| Content | Whether pages answer questions directly and name things unambiguously | Review against a prompt set and, for Google, comparison with the AI Overview a page competes with |
| Reputation | What other sites, forums and review platforms say about you | The sources AI tools cite for your prompts, brand mentions and review profiles |
| Product data | Whether page, markup and feed agree on price, availability and attributes | Feed validation, structured data testing and Merchant Center diagnostics |
The first two rows are usually the cheapest to fix, and a problem there can undermine the work in the others, so that is where I start. If you sell products, the ecommerce SEO page covers product data and AI shopping in more depth.
An illustrative example. This is a typical pattern, not a client case. A retailer may find it is missing from AI shopping answers even though its products rank well organically. The investigation might show that key product attributes are missing from the feed, that some product details only appear after JavaScript runs, and that AI answers lean on third-party sites carrying inconsistent product information.
Those are three different problems with three different owners. Rewriting the product page copy would fix none of them.
Tools I built for this work
I build a tool when I find myself answering the same question repeatedly. These are the ones relevant to AI search, and all of them are free:
AI Visibility Prompt Generator
Turns a page's real content into the questions people would plausibly ask an AI tool, giving you a prompt set to test and track.
Content vs AI Overview comparison
Scores how closely a page aligns with the AI Overview it competes with and shows which phrases are missing. Runs in Google Colab; the cosine similarity guide explains it.
AI Product Feed Readiness Checker
Validates a product feed against Google Merchant Center requirements, including the attributes that matter for conversational shopping.
Organic Trends
Separates brand from non-brand performance using Search Console and GA4, so you can see where a change in organic traffic is actually happening.
PLP & PDP FAQ Generator
Drafts category and product FAQs from search data, AI Overviews and Reddit discussions.
Having built them is also why I am careful about what they can tell you. A tool shows a signal. It does not show why an AI system chose one source over another.
Measuring AI visibility honestly
This is the weakest part of most GEO proposals, so it is worth being precise about what can and cannot be measured.
- AI referral traffic. Visits from chatgpt.com, perplexity.ai, gemini.google.com and similar sources can be separated in GA4 with a custom channel. It undercounts, because many people read an answer and never click, or search for you by name later.
- Google's AI features. Search Console includes AI Overviews and AI Mode within its normal web search reporting rather than breaking them out, so their effect has to be inferred from query-level changes.
- Prompt tracking. Running a fixed set of prompts on a schedule, across several AI tools, shows whether you are mentioned or cited and which sources appear instead. Because answers vary, the trend across the whole set is meaningful; a single result is not.
- Server logs. Logs show which AI crawlers request which pages, and how often. User-agent names can be faked, so requests are verified against the IP ranges or reverse DNS lookups the crawler operators publish. That confirms access rather than visibility, but it catches blocks nothing else reveals.
- Brand search. Movement in searches for your name is a supporting metric only. Brand demand has too many causes to attribute it to AI exposure on its own.
I will tell you when the data cannot yet show whether something worked.
What I will not sell you
- Guaranteed citations or placements in any AI answer. Nobody controls those.
- A special file, markup or schema type presented as the key to AI visibility.
- A rewrite of every page “for AI” when the real problem is access, indexing or reputation.
- Fake reviews, planted forum posts or undisclosed promotion on Reddit. They break platform rules and tend to damage the reputation they were meant to build.
What you receive
A typical AI search review includes:
- A crawler access assessment covering robots.txt, the relevant AI user agents, and CDN or firewall restrictions.
- An agreed prompt set, built from real customer questions and commercial search intent.
- A visibility baseline showing mentions, citations, how you are described and which competing sources appear, across the AI platforms that matter to your audience.
- A prioritised roadmap, with each recommendation assigned to the team best placed to implement it.
- A measurement framework that keeps work delivered, AI visibility, referral traffic and commercial outcomes separate.
- A playback session with your marketing, content, ecommerce and technical teams.
How this usually starts
With a review in four steps, before any ongoing work is agreed.
-
Check access
Robots.txt, CDN and firewall rules, server logs and rendering for your most important page templates. It is the quickest layer to check, so it comes first.
-
Build the prompt set
The questions your customers actually ask, taken from Search Console queries, site search, sales and support conversations and your own content, grouped by intent and stage of decision.
-
Baseline
The prompt set run across the AI tools that matter to your audience.
-
Prioritise
Fixes on your site, content work, and off-site work that needs PR or community input, ordered by likely impact.
A review is scoped and quoted as a fixed-price project. Where you want prompt tracking and ongoing changes, it can continue as a monthly retainer. How my pricing works is set out on its own page.
Frequently asked questions
Is GEO different from SEO?
Mostly it extends it. Crawlable, indexed, clearly written pages remain the foundation. GEO adds attention to AI crawler access, third-party reputation, how AI tools describe you, and measuring visibility in answers as well as rankings.
Can you guarantee my business will appear in ChatGPT or AI Overviews?
No. AI systems choose their own sources, and answers vary between users and sessions. What can be done is removing obstacles and measuring whether visibility changes across a consistent set of prompts.
Should we block AI crawlers?
It depends on what you want. Blocking training crawlers is a reasonable choice for some publishers, but blocking a platform's search crawler can prevent or significantly reduce the use of your pages in its answers. Google-Extended does not affect inclusion in Google Search, including AI Overviews. Decide deliberately, crawler by crawler.
How long before GEO work shows results?
It varies with the problem. A crawler block can be fixed quickly, although AI tools still have to revisit the pages. Content changes move at roughly the pace of search, and changing how other sites discuss you is slow. I set expectations for each piece of work rather than give one number.
Do you only work with businesses in London?
No. I am based in London and work with businesses across the UK and internationally, mostly remotely, with in-person meetings in London where useful.
Where to start
I keep the number of consultancy clients small, so the checks and the judgement behind them stay with me rather than being passed to someone junior.
If you would rather test the thinking first, read why being cited now matters as much as being ranked, or run one of your pages through the AI Visibility Prompt Generator.
Otherwise, send me three things: your site, the questions you want to be the answer to, and anything you have seen AI tools say about you so far. I will tell you what I would check first, and whether consultancy is worth paying for at this stage.