Specialist SEO
AI Search Optimisation That Makes Your Business the Source Answers Cite
More of your buyers now ask an AI assistant before they ask Google, and the answer they get names a handful of businesses. I work on the things that make those systems able to find, understand and quote you, and I am honest about what nobody can control.
- years in SEO
- 6+
- client websites worked on
- 60+
- person doing the work
- 1
AI search optimisation for when the answer arrives before the click
AI search optimisation is the work of getting your business found, understood and cited in the answers AI Overviews, ChatGPT, Perplexity, Copilot and Gemini give your buyers. Ask Google a question and you may see an AI Overview above the results; ask a chat assistant and you get a written answer with a few sources linked underneath, or none at all. For a growing share of research-stage searches, that answer is the first impression of your market, and the businesses named in it start the shortlist.
These systems are not ranking pages in the old sense. They retrieve a set of documents, summarise them, and attribute some of the claims. Which documents get retrieved and which get credited is not published in detail by any of the companies involved, and it changes often. What I can describe are the patterns: what pages that get cited tend to have in common, and what stops a site from being considered at all.
That is the honest frame for this work, sometimes sold as generative engine optimisation. It is not a hack. It is making your site technically reachable, your business unambiguous, and your content the clearest available answer to the questions your buyers ask.
Key takeaways
- Most AI search products retrieve live pages from a search index before writing an answer, so a page that is not crawled and indexed cannot be cited, however good it is.
- Clear entity information - who you are, what you do, where and for whom - repeated consistently on your site and elsewhere makes you easier to name in an answer.
- Pages that answer a specific question directly, near the top of a section, tend to be quoted more often than pages that bury the answer in an introduction.
- Nobody can guarantee a citation. Answers vary by user, wording and session, so I measure trends across many prompts and sources rather than single screenshots.
- Most of the work overlaps with good SEO. The difference is in emphasis: answer-shaped content, corroboration from other sites, and crawler access decided deliberately.
What's included in my AI search optimisation
Seven parts, in proportions set by the baseline: a site Bing barely indexes needs access work first, while a well-indexed site with generic copy needs content and corroboration.
A tracked prompt set and a baseline of who AI answers name today
With you, I list the questions buyers ask at each stage - definitions, comparisons, costs, provider shortlists - in their own words. I run them across several AI products and record where you are named, which competitors and sources appear instead, and what the cited pages have in common.
- A prompt set agreed before work starts
- Baseline across Google, Bing-based and chat products
- Patterns in the pages that get cited
Crawler and index access decided on purpose, not copied from a forum
A deliberate robots.txt policy that separates search crawlers from training crawlers where vendors allow it, set crawler by crawler from current documentation. Bing indexing is checked rather than assumed, because Copilot depends on it, and nothing important stays hidden behind scripts or logins.
- Robots.txt policy for each AI crawler
- Indexing review in Google and Bing
- Rendering checks on key pages
Entity clarity so every source describes your business the same way
One consistent description of who you are, what you do, where and for whom, across the site, Business Profile, directories and social profiles. That includes an About page that says it plainly, Organization schema and links to the profiles that confirm the details.
- Entity consistency audit with fixes listed
- About page and Organization schema
- Profile links that corroborate each other
Answer-shaped content written so a passage can be quoted on its own
Question-led headings that match how buyers phrase things, a direct answer in the first sentences of each section, then the detail. Definitions, comparisons, costs and processes are written to make sense when lifted out. Existing pages get rewritten first; new ones are created only where nothing fits a prompt cluster.
- A prioritised list of pages to rewrite or create
- The prompts each page should answer
- Direct answers near the top of each section
Named authors and first-hand detail that give a system a reason to cite you
Content attributed to real people with visible credentials, specific first-hand detail and dated updates. Generic text that could sit on any competitor's site gives an answer engine no reason to pick you, so I work with your experts to add what only your business knows.
- Author pages with real credentials
- First-hand examples and specifics
- Visible update dates on key pages
Corroboration on the third-party sources AI answers already cite
Mentions on the industry sites, directories, review platforms and publications that answers in your sector already reference. I find them by checking which domains appear in answers for your prompts, then go after the ones where a listing or mention is realistic.
- A target list of cited domains in your sector
- Listings and editorial mentions pursued
- Review requests within each platform's rules
Measurement that combines prompt tracking, AI referrals and citation data
There is no AI equivalent of Search Console, so I combine imperfect signals, set out in the table below, and read them as a trend. If you want hundreds of prompts tracked across platforms with a screenshot per answer, that runs as a pipeline under AI search visibility tracking.
- A fixed prompt set rerun on a schedule
- AI referral traffic separated from other referrals
- Citation data from the platforms that report it
Experience
AI Search 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
The AI surfaces I work on and how they tend to find sources
Each product works differently and none publishes its full method. This is what I work from, based on the vendors' own documentation and what I observe on client sites.
| Surface | Where sources tend to come from | What I focus on |
|---|---|---|
| Google AI Overviews and AI Mode | Google's own search index; Google says a page needs to be indexed and eligible to show a snippet | Indexing, snippet controls, and pages that answer the question in a self-contained passage |
| Microsoft Copilot | Bing's index, which many site owners neglect | Bing Webmaster Tools, sitemap submission, Bing indexing and any citation data it reports |
| ChatGPT search | Live retrieval through OpenAI's search crawler and search partners, plus what the model already knows | Robots.txt rules for OpenAI's search and training crawlers set separately and on purpose |
| Perplexity | Live web retrieval with sources shown prominently | Crawler access, fast and clean pages, and clear claims attributed to named people |
| Gemini | Google's index and knowledge systems | The same foundations as Google Search, plus entity consistency across profiles |
Crawler names, controls and reporting change. I check each vendor's current documentation at the start of an engagement rather than relying on last year's list.
What gets sold as AI optimisation, and what I do instead
The category is new, so the shortcuts are everywhere. Some of them carry real risk.
Common shortcuts
- Hundreds of AI-written pages targeting every possible question
- Bolted-on FAQ sections stuffed with brand mentions
- Paid placements on list articles that exist only to be scraped
- Fake or incentivised reviews to look popular
- Dashboards claiming a share-of-voice score from a dozen prompts
- Special files or markup presented as a guaranteed route into AI answers
My approach
- Fewer, better pages that answer the questions buyers actually ask
- FAQs only where they help the reader, written in plain language
- Earned mentions on sources that already appear in answers for your topic
- Review requests at the right moment, within each platform's rules
- Large, stable prompt sets read as a trend, with their limits stated
- Standards adopted when there is evidence they are used, not because they are fashionable
Free review
Get a free AI Search 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.
What this is not
Not a separate discipline that replaces SEO
If your site is not indexed properly, AI search work will not help, and that is a job for technical SEO. If you have no plan for what to publish, that comes from content strategy, which this page builds on rather than replaces. Brand mentions, reviews and wider reputation sit under off-page SEO. AI search optimisation is the layer that shapes all three around how answer engines retrieve and quote information.
How an AI search engagement runs
The first baseline and plan take around three to four weeks. Content and corroboration work then runs month by month.
- 1
Build the prompt set
First week, in a working session with you. The set then stays fixed so later runs can be compared like for like.
- 2
Take a baseline across surfaces
Weeks one to three, with each prompt run more than once per product, because a single run of any prompt can mislead.
- 3
Check access and entity signals
In parallel with the baseline, since access problems block everything else and are often the quickest wins.
- 4
Close the answer gaps
From month two, cluster by cluster, starting with the prompts closest to a buying decision.
- 5
Earn corroboration
Runs alongside the content work, aimed at the sources the baseline showed being cited for your topics.
- 6
Re-measure and adjust
On a regular schedule, with effort moved towards the prompt clusters and sources that respond.
How AI visibility gets measured, and the limits of each signal
There is no single AI equivalent of Search Console. I combine several imperfect signals and report them side by side.
| Signal | Where it comes from | What it shows | Its limit |
|---|---|---|---|
| Prompt tracking | A fixed set of prompts run on a schedule, by hand or with a tool | Whether you are named or cited, and who is named instead | Answers vary by person, location and session; read it as a trend |
| AI referral traffic | Analytics sessions from domains such as chatgpt.com, perplexity.ai and Copilot or Gemini referrers | Visits and enquiries that arrive from AI answers | Many answers send no click, and some visits arrive without a referrer |
| Platform citation data | Reports from search engines that expose AI citations, such as Bing Webmaster Tools | How often your pages are referenced in that product's answers | Covers one product only, and definitions differ by platform |
| Search Console | Google's performance report, where AI features are counted within web search | Overall Google visibility for the queries AI Overviews appear on | AI Overview impressions are not reported separately |
| Enquiry source | A "how did you hear about us" field and sales call notes | Buyers who say an AI assistant recommended you | Self-reported and incomplete, but often the most persuasive evidence |
After the click
Visitors from AI answers arrive already half-convinced
Someone who clicks through from an AI answer has usually read a summary of your offer already. They want confirmation, not an introduction: proof that the claim they read is true, pricing or a next step, and a quick way to talk to you. I check the pages AI answers send traffic to and make sure the first screen matches what the answer said about you. A mismatch between the summary and the page is one of the fastest ways to lose these visitors.
Common problems and where this fits
The issues I find most often are mundane. Robots.txt rules copied from a forum that block the search crawlers a business wants while allowing the ones it wanted to block. Sites indexed well in Google and barely at all in Bing, which matters for Copilot. Service pages that never state plainly what the business does or where. Authors with no name. And three slightly different company descriptions across the site, Business Profile and LinkedIn.
For examples of this work in practice, see question-led content for a private-markets investor club, and opening up a batch of hidden pages for a language services company. Both were measured with the kind of citation reporting described above.
AI search connects to almost everything else I do. Indexing problems go to technical SEO, the editorial plan to content strategy, mentions and citations to link building and off-page SEO, and multi-language visibility to international SEO. If you want everything checked at once, start with an SEO audit.
Growth from real projects
Learn more about ai search optimisation
AI Search Optimisation FAQs
Is AI search optimisation different from SEO?
Mostly it is SEO with a different emphasis. The foundations are the same: crawlable, indexed pages, a clear business identity and content that answers real questions. The difference is in how content is shaped, so a passage can be quoted on its own, and in paying more attention to third-party sources and Bing, because several AI products draw on them.
Can you guarantee we will appear in ChatGPT or AI Overviews?
No, and you should be wary of anyone who does. The companies behind these products do not publish how sources are chosen, answers change between users and sessions, and the systems are updated often. What I can do is remove the obstacles, make your content the clearest answer available and measure whether citations and AI referrals move over time.
Should we block AI crawlers in robots.txt?
It depends on what you want. Some vendors run separate crawlers for search features and for model training, so you may be able to allow one and block the other. Blocking a search crawler can remove you from that product's answers. I set the policy with you crawler by crawler, based on each vendor's current documentation, rather than copying a blanket rule.
Does llms.txt or special schema get us into AI answers?
I have not seen solid evidence that the major AI search products rely on an llms.txt file, so I treat it as optional and low priority. Structured data such as Organization schema helps describe your business consistently, which supports entity clarity, but no markup guarantees a citation. Content quality, crawl access and corroboration from other sites matter far more.
How long before we see changes in AI answers?
Fixes to crawl access and indexing can show within weeks, once pages are recrawled and retrieved. Changes driven by new content and third-party mentions take longer, usually a few months, and progress is uneven across products. I track a fixed prompt set and citation data from the start, so you can see the trend rather than judge on a single answer.
Which businesses benefit most from AI search work?
Businesses whose buyers research before contacting anyone: professional services, B2B suppliers, software, healthcare and specialist retail. If your customers ask questions like which provider to choose, what something costs or how a process works, AI answers are already shaping their shortlist. Purely local emergency services usually gain more from map pack work first.
Get a free AI search optimisation check for your market
Send me five questions your buyers ask before they get in touch. On a free 30-minute call I will show you who AI answers name for them today, which sources they cite, and the first changes I would make to get your business into those answers.