AI Search Optimisation: How to Get Your Website Cited by ChatGPT, Google AI Overviews, Gemini, Perplexity and Claudetitled

Search is changing.

People are no longer only typing short queries into Google and working their way through a list of blue links. They are asking ChatGPT, Gemini, Claude, Perplexity and Microsoft Copilot complete questions such as:

• Which accounting software is best for a small UK ecommerce business?

• What is the best virtual waiting room platform for a ticketing website?

• Which growth marketing agencies specialise in startups?

• What should I consider before choosing a private healthcare provider?

• Compare the leading CRM platforms for a ten-person sales team.

The AI system then researches the subject, evaluates information from multiple sources and gives the user a summarised answer.

Sometimes it mentions a business.

Sometimes it recommends a product.

Sometimes it cites and links to the website it used.

And sometimes a company that ranks well in traditional search results is missing entirely.

That has created a new marketing discipline commonly described as:

• AI search optimisation

• Generative Engine Optimisation, or GEO

• Answer Engine Optimisation, or AEO

• LLM optimisation

• AI visibility optimisation

The terminology matters less than the objective:

Make your business and its content easy for AI systems to discover, understand, trust,

retrieve and cite.

This guide explains how.

What is AI search optimisation?

AI search optimisation is the process of improving your website, content and wider online presence so that AI-powered search systems are more likely to:

• find your content

• understand what your business does

• recognise your expertise

• retrieve your pages for relevant questions

• use your information when constructing an answer

• mention your brand

• cite and link to your website

• recommend your products or services where appropriate

Traditional SEO normally focuses on earning a visible position in a ranked list.

AI optimisation focuses on earning inclusion within a generated answer.

That changes the format of the opportunity, but it does not make SEO obsolete.

Google explicitly says its generative search features remain rooted in its core Search ranking and quality systems. AI Overviews and AI Mode use techniques including retrieval-augmented generation and “query fan-out”, where several related searches may be performed to gather enough information to answer a more complex question.

The practical conclusion is straightforward:

Good AI optimisation normally begins with good SEO. It then adds stronger content, clearer entity signals, better evidence and broader authority across the web.

How AI search systems find information

It is easy to imagine that ChatGPT or Gemini simply “knows” everything contained in its answer.

That is not how modern AI search always works.

An answer may draw on two broad sources of information.

1. Information learned during model training

Large language models are trained on substantial collections of information. This gives them broad knowledge of established subjects, major businesses, products, people and concepts.

However, training data has limitations:

• it may not include your business

• it may contain old information

• it may reflect conflicting sources

• the model may not be able to identify precisely where a fact originated

• a recently published page may not have been included

This means being present in potential training data is not the same as appearing in live AI search results.

2. Information retrieved from the live web

Many AI products can search or retrieve web content when answering a question.

The system may:

1. interpret the user’s question

2. generate related searches

3. retrieve potentially useful pages

4. extract relevant passages

5. compare information from different sources

6. generate an answer

7. cite some of the supporting pages

Google describes this as retrieval-augmented generation, or RAG. Its “query fan-out” process may run several related searches alongside the original query.

Claude’s web search can also conduct searches and return answers with cited sources, while Perplexity operates a continuously refreshed search index and produces citation-led answers.

Your website therefore has to succeed at several different stages.

The four stages of AI visibility

1. Discovery

Can the search engine or AI crawler access and find the page?

2. Retrieval

Does the system consider the page relevant to the user’s question?

3. Extraction

Can it clearly identify the useful facts, explanation, recommendation or evidence?

4. Selection

Does it trust the information enough to use or cite it instead of competing sources?

AI optimisation has to address all four.

How the major AI platforms access websites

There is no single universal AI index.

Different products rely on different crawlers, search indexes and retrieval systems.

ChatGPT

OpenAI operates a crawler called OAI-SearchBot for ChatGPT search. OpenAI says websites that opt out of this crawler will not be included as sources within ChatGPT search answers, although a page may still occasionally appear as a basic navigational link.

OpenAI separates this from GPTBot, which is associated with potential model training.

You can therefore allow your website to appear in ChatGPT search while separately blocking the use of your content for model training. The two controls are independent.

Google AI Overviews, AI Mode and Gemini

Google’s generative search experiences draw on pages in the Google Search index.

To be eligible, a page needs to be crawlable, indexed and permitted to appear with a search snippet. Google says its existing technical and content SEO guidance continues to apply.

Microsoft Copilot

Microsoft’s web-grounded Copilot systems can search public websites indexed by Bing. This makes Bing indexing more important than many businesses previously assumed.

Perplexity

Perplexity operates PerplexityBot, which it describes as a crawler designed to surface and link websites within Perplexity search results. It is separate from user-triggered visits and is not used to crawl content for foundation-model training.

Claude

Anthropic operates several separate bots.

Claude-SearchBot is used to improve the relevance and accuracy of web search results.

ClaudeBot relates to potential model training, while Claude-User supports visits initiated by individual Claude users.

Again, website owners can make different decisions about search visibility and model training.

How to optimise your website for AI search

There is no switch you can activate to guarantee citations.

There is also no reliable collection of hidden prompts, special files or schema tags that will suddenly make an AI system recommend your company.

The strongest approach is to improve the underlying information environment around your business.

That means working across five areas:

1. Technical accessibility

2. Content quality and extractability

3. Business and entity clarity

4. External authority and corroboration

5. Measurement and continuous improvement

Let’s look at each one.

1. Make sure AI systems can access your website

A brilliant guide cannot be cited if the systems producing the answer cannot retrieve it.

Check your robots.txt file

Review the robots.txt file located at:

yourwebsite.com/robots.txt

Make sure you have not accidentally blocked the search crawlers you want to access your content.

A permissive configuration might include:

User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

You do not usually have to add an explicit Allow rule if your general robots.txt configuration already permits crawling. The important point is to check that these bots are not being disallowed elsewhere.

Decide separately whether you want to allow training-related crawlers such as GPTBot or ClaudeBot.

Blocking a training crawler does not necessarily require you to block the corresponding search crawler.

Check your firewall and bot protection

Robots.txt is only one layer.

Cloudflare, hosting security tools, WordPress security plugins and web application firewalls may block unfamiliar bots even when robots.txt allows them.

Review:

• firewall events

• server logs

• CDN bot settings

• rate-limiting rules

• managed challenges

• CAPTCHA settings

• country or IP restrictions

Perplexity specifically advises websites using a WAF to verify requests through both the official user agent and its published IP ranges.

Do not blindly whitelist anything claiming to be an AI crawler. Verify it against the platform’s official documentation and published IP data.

Make important content available as visible HTML

Your core information should not exist only inside:

• images

• videos

• downloadable PDFs

• interactive widgets

• tabs that fail to render

• content loaded only after a complex user action

• scripts that some crawlers cannot execute

• forms requiring personal information

• logged-in areas

Google can render JavaScript, but JavaScript-heavy websites are still more complex to crawl and diagnose. Google recommends following its normal JavaScript SEO guidance and ensuring that important content is accessible.

As a practical rule, view the page source or disable JavaScript temporarily.

Can you still see the page’s primary heading, description, service information, product details and main evidence?

If not, investigate server-side rendering, static rendering or progressive enhancement.

Resolve basic indexing problems

Check that important pages:

• return a 200 status code

• are not marked noindex

• are not blocked by robots.txt

• have the correct canonical URL

• are included in your XML sitemap

• do not redirect unnecessarily

• are not duplicated across several URLs

• are internally linked

• are not orphaned

• load reliably

• work on mobile devices

Use both Google Search Console and Bing Webmaster Tools.

Bing visibility deserves particular attention because Microsoft’s web-grounded Copilot experiences can use websites indexed by Bing. Bing’s IndexNow system can also notify participating search engines when URLs are created, changed or removed.

Improve page accessibility

Clear page structure benefits people, search engines and AI agents.

Use:

• one clear primary heading

• logical H2 and H3 subheadings

• descriptive link text

• labelled buttons

• meaningful image alt text

• properly associated form labels

• accessible menus

• clear ARIA labels where necessary

OpenAI says its agents use accessibility information, including ARIA roles and labels, to interpret interactive elements.

Accessibility is therefore becoming part of both user experience and machine usability.

2. Create content worth citing

Most websites do not have a technical AI problem.

They have an information problem.

Their pages say roughly the same things as every competitor:

• We offer tailored solutions.

• We put customers first.

• We are passionate about quality.

• Contact us for more information.

• Here are five generic tips you could find anywhere.

There is very little for an AI system to cite.

Google’s latest guidance places particular emphasis on what it calls “non-commodity content”: information that adds something distinctive rather than restating common knowledge. It recommends first-hand experience, original viewpoints and information that could not be easily generated by summarising existing pages.

Answer the question early

Do not make readers search through 500 words of scene-setting to find the answer.

A strong informational page should normally tell the reader:

• what the subject is

• who the advice is for

• the direct answer

• the important qualifications

• what to do next

Put a concise answer near the beginning, then expand on it.

For example:

A virtual waiting room is a system that places excess visitors into an online queue when demand exceeds a website’s safe capacity. It is commonly used for ticket sales, product launches and limited releases to prevent crashes, bots and overselling.

That passage is far easier to retrieve and cite than:

In today’s fast-moving digital landscape, businesses face an unprecedented range of challenges when trying to deliver seamless digital experiences.

The second version says almost nothing.

Write around complete customer questions

Traditional keyword research may show a term such as:

CRM for startups

But AI users are more likely to ask:

• What CRM is best for an early-stage B2B startup?

• Is HubSpot or Pipedrive better for a five-person sales team?

• What is the cheapest CRM that still includes automation?

• Which CRM is easiest to migrate away from later?

• Do we need a CRM before hiring our first salesperson?

• What CRM integrates with our billing and customer support tools?

Build your content around the underlying decisions, not just a two-word keyword.

Anticipate query fan-out

An AI system may break a question into several supporting questions.

Someone asking, “Which email marketing platform should I use?” may trigger research into:

• pricing

• contact limits

• automation features

• ecommerce integrations

• deliverability

• ease of use

• support

• migration

• data protection

• suitability for different business sizes

A page that answers only one of these may be useful.

A well-structured guide that covers the complete decision is more useful.

This does not mean creating hundreds of thin pages for every wording variation. Google specifically warns against producing separate low-value pages simply to target large numbers of fan-out or long-tail queries.

Cover the subject properly rather than producing more pages for the sake of volume.

Add original evidence

The most valuable content contains information that other websites cannot simply reproduce without citing you.

Examples include:

• original research

• anonymised customer data

• industry benchmarks

• survey results

• pricing analysis

• experiments

• performance data

• proprietary frameworks

• first-hand product tests

• expert commentary

• detailed case studies

• implementation examples

• calculators

• templates

• datasets

• before-and-after comparisons

Instead of writing:

Improving your website can increase conversions.

Write:

After removing two unnecessary fields from the application form, completion increased from 31% to 44% across 6,200 sessions.

Include:

• what was changed

• the dates of the test

• the sample size

• the metric used

• any limitations

• the final result

That gives the reader and the AI system a precise, attributable fact.

An early academic study into Generative Engine Optimisation found that adding credible citations, relevant statistics and clear supporting evidence could improve source visibility within the experimental generative engines tested. The researchers reported gains of up to 40% in some circumstances, although performance varied considerably by subject and this should not be treated as a universal ranking formula.

The sensible lesson is not to insert random statistics.

It is to make useful claims verifiable.

Include clear comparisons

Comparison content is particularly valuable because AI users frequently ask for help making decisions.

Strong comparison pages should explain:

• which options are being compared

• the criteria used

• who each option suits

• important differences

• disadvantages as well as advantages

• price or commitment differences

• the evidence behind the conclusion

• when neither option is suitable

Avoid declaring your own product the winner in every category.

A visibly biased comparison is less credible to people and offers weaker evidence to an answer engine.

Include limitations and exceptions

Trustworthy pages explain where their advice does not apply.

Use statements such as:

• This approach works best when…

• This may not be appropriate if…

• The main disadvantage is…

• Results will depend on…

• This figure excludes…

• We would not recommend this for…

• There is not yet enough evidence to conclude…

AI systems often have to reconcile conflicting information. A page that recognises nuance is usually more useful than one written entirely in absolute marketing claims.

Show who created the content

For important guides, include:

• a named author

• a useful author biography

• relevant experience

• links to the author’s profile

• an original publication date

• a “last reviewed” or updated date

• editorial or review information where appropriate

• sources for material factual claims

A generic “Admin” author page does little to establish expertise.

A biography should explain why the person is qualified to write about that specific subject.

Keep important content current

AI search often favours fresh information for questions involving:

• prices

• products

• laws

• software

• platform features

• statistics

• recommendations

• industry trends

• medical or financial guidance

• business comparisons

Review important pages on a defined schedule.

Do not simply change the date.

Check:

• whether the answer remains correct

• whether products have changed

• whether quoted prices still apply

• whether external links still work

• whether statistics have newer equivalents

• whether recommendations need revising

• whether screenshots reflect the current interface

Meaningful updates create value. Artificially refreshing the date does not.

3. Make your business easy to understand as an entity

AI systems need to understand not only your page, but also the organisation behind it.

They may encounter your business across:

• your website

• LinkedIn

• Companies House

• Google Business Profile

• review platforms

• trade associations

• partner websites

• news articles

• directories

• podcast pages

• event listings

• customer case studies

• social profiles

Conflicting information creates ambiguity.

Use one consistent business identity

Keep the following consistent wherever possible:

• business name

• trading name

• website address

• logo

• short description

• address

• service area

• telephone number

• founder names

• social profiles

• categories

• founding date

• products and services

This is particularly important when:

• the business has changed name

• the domain differs from the company name

• several similarly named organisations exist

• the business uses multiple regional websites

• an old product or brand remains widely referenced

• the company operates under a trading name

Create a clear About page that states precisely:

• who you are

• what you provide

• who you work with

• where you operate

• what makes your model different

• who leads the business

• how long you have operated

• how customers can contact you

Add appropriate structured data

Structured data gives search systems explicit information about the meaning of a page.

Useful types may include:

• Organization

• LocalBusiness

• Person

• ProfilePage

• Article

• Product

• Offer

• Service

• BreadcrumbList

• Event

• JobPosting

• Dataset

Google recommends JSON-LD where practical and says structured data can help it understand the people, organisations, products and other entities described on a page.

For an organisation, useful properties can include:

• official name

• URL

• logo

• description

• address

• telephone number

• email address

• company identifiers

• relevant social and profile URLs through sameAs

Google says Organisation structured data can help it disambiguate a company and understand administrative details.

However, structured data is not an AI visibility cheat code.

Google explicitly says there is no special schema required for generative AI search and that structured data is not mandatory for inclusion. It should support the information already visible on the page, not replace it.

Connect experts to subjects

If your business relies on specialist knowledge, create strong profile pages for your experts.

Each page can include:

• name and role

• areas of expertise

• experience

• qualifications

• relevant clients or projects

• articles written

• conference appearances

• interviews

• professional memberships

• external profiles

Then connect each article or guide to the relevant expert.

This creates a much clearer relationship between:

Person → expertise → organisation → content → evidence

4. Build authority beyond your own website

Your website is controlled by you.

An AI system knows that.

It may therefore look for independent confirmation of what you say.

Your site may claim you are the UK’s leading specialist in a category. That claim becomes more credible when it is supported by:

• customers

• respected publications

• industry bodies

• professional directories

• review platforms

• recognised experts

• independent comparison websites

• business partners

Earn relevant third-party mentions

Useful coverage may include:

• digital PR

• expert commentary in news articles

• industry publication contributions

• original research cited by journalists

• conference speaker profiles

• podcast guest pages

• supplier and partner directories

• award listings

• professional association profiles

• university or public-sector references

• customer case studies

• integration marketplaces

The objective is not to distribute the same promotional press release across dozens of low-quality sites.

It is to create independent, relevant and accurate evidence that connects your brand to the subjects on which you want to be recognised.

Give other people something worth referencing

Generic opinion pieces rarely attract meaningful coverage.

Reference-worthy assets include:

• annual industry reports

• proprietary datasets

• market maps

• interactive tools

• calculators

• detailed benchmarks

• public templates

• strong visualisations

• expert-led predictions

• transparent experiments

• well-documented case studies

The asset should live on your website, while outreach helps relevant third parties discover and cite it.

That strengthens both direct visibility and the wider evidence network around your brand.

Strengthen your review presence

Reviews can help answer engines understand:

• what customers use you for

• common strengths and weaknesses

• the types of customer you serve

• product quality

• service reliability

• recurring themes in customer experience

Focus on the platforms that matter within your market.

These might include:

• Google

• Trustpilot

• G2

• Capterra

• Clutch

• Feefo

• industry-specific directories

• marketplace reviews

Do not manufacture reviews or ask customers to insert unnatural keywords.

Ask for honest, specific feedback about:

• the problem

• the service provided

• the outcome

• who the service is suitable for

Participate in genuine industry conversations

Forums, communities and social platforms can influence how a brand is discussed online.

That does not mean filling Reddit, Quora or LinkedIn with disguised advertising.

It means:

• answering genuine questions

• correcting misinformation

• sharing relevant experience

• explaining trade-offs

• disclosing your affiliation

• contributing before promoting

A discussion in which independent users recommend your product is more persuasive than ten posts written by newly created brand accounts.

5. Build topical depth, not a random blog

Publishing one article about a subject does not automatically establish expertise.

Your website should demonstrate that you understand the wider problem your customer is trying to solve.

Create a topic and question map

Begin with your main commercial themes.

For each theme, map questions across the buying journey.

Problem awareness

• Why does this problem happen?

• What are the risks?

• How serious is it?

• What happens if it is ignored?

Solution awareness

• What options exist?

• How does each solution work?

• Can the problem be solved internally?

• What are the alternatives?

Evaluation

• Which solution is best for a particular use case?

• What features matter?

• What does it cost?

• What are the drawbacks?

• How long does implementation take?

Purchase

• What is included?

• What support is available?

• What happens during onboarding?

• What information is needed to begin?

Post-purchase

• How do I implement it?

• How do I measure success?

• What problems are common?

• How do I get better results?

This creates a connected information resource rather than a collection of unrelated articles.

Connect informational and commercial pages

Your guides should link naturally to:

• relevant service pages

• product pages

• case studies

• comparison pages

• tools

• templates

• expert profiles

Your commercial pages should also link back to supporting evidence.

For example, a service page can connect to:

• a complete guide explaining the problem

• a case study showing the outcome

• a methodology page explaining your process

• an FAQ addressing objections

• an expert profile establishing credibility

Internal linking helps search systems discover these relationships and helps users continue their research.

How different types of businesses should approach AI optimisation

The fundamentals are shared, but priorities differ by business model.

Service businesses

Prioritise:

• detailed service pages

• clear sector specialisms

• named experts

• case studies

• pricing guidance where possible

• process explanations

• comparison pages

• geographic coverage

• reviews

• expert commentary on third-party sites

Avoid vague pages that simply repeat the service name followed by generic promises.

SaaS companies

Prioritise:

• feature documentation

• integration pages

• use-case pages

• implementation guides

• transparent pricing

• comparison pages

• migration guides

• security documentation

• customer evidence

• changelogs

• product terminology consistency

AI users often ask very specific product questions. Your documentation may be as important as your marketing pages.

Ecommerce businesses

Prioritise:

• accurate product data

• complete product descriptions

• availability

• pricing

• variants

• delivery information

• return policies

• product identifiers

• reviews

• comparison content

• high-quality images

• Merchant Center feeds

• product structured data

Google recommends combining accurate product structured data with Merchant Center information to improve its understanding of ecommerce products and eligibility across shopping surfaces.

Retailers can also submit product feeds for discovery within ChatGPT shopping experiences.

Local businesses

Prioritise:

• Google Business Profile

• consistent name, address and telephone information

• service-area information

• opening hours

• local landing pages

• genuine local reviews

• photographs

• LocalBusiness structured data

• relevant local directories

• local press and community references

Google says Business Profiles and appropriate local business information can help businesses appear within both traditional and generative Search experiences.

Publishers and expert-led brands

Prioritise:

• author profiles

• editorial standards

• source citations

• original reporting

• clear dates

• corrections policies

• topic expertise

• primary documents

• original research

• strong internal linking

The goal is to become the source other pages summarise rather than another page summarising everyone else.

How to measure AI visibility

AI optimisation is more difficult to measure than conventional rankings.

Answers can vary according to:

• the wording of the prompt

• the user’s location

• the platform used

• whether web search is active

• the date

• model updates

• personalisation

• follow-up questions

• which sources were recently indexed

You therefore need several forms of measurement.

Track AI referral traffic

In analytics, monitor referral traffic from domains associated with:

• ChatGPT

• Perplexity

• Gemini

• Claude

• Copilot

• Bing

• other emerging answer engines

OpenAI automatically adds utm_source=chatgpt.com to referral URLs from ChatGPT search, making this traffic easier to identify.

Track more than sessions.

Measure:

• engaged sessions

• landing pages

• lead quality

• registrations

• purchases

• assisted conversions

• revenue

• conversion rate

• time to conversion

AI referral traffic may be lower in volume but higher in intent because the user has already completed part of their research.

Use Google Search Console’s AI reporting

In June 2026, Google announced dedicated Search Console reporting for visibility within generative AI features such as AI Overviews, AI Mode and generative experiences in Discover.

The reports can show:

• impressions

• pages displayed

• countries

• devices

• performance over time

The feature was initially being rolled out to a subset of websites, so it may not yet appear in every Search Console account.

Monitor crawler activity

Review server or CDN logs for visits from:

• OAI-SearchBot

• PerplexityBot

• Claude-SearchBot

• Bingbot

• Googlebot

This will not tell you whether the page was cited, but it can reveal:

• whether the crawler can access the site

• which pages are being visited

• repeated errors

• blocked requests

• crawl frequency

• sections receiving little attention

Build a prompt monitoring set

Create a controlled list of important questions.

Include:

• category questions

• comparison questions

• recommendation questions

• problem-solving questions

• brand questions

• pricing questions

• local questions

• use-case questions

For example:

• What are the best growth marketing agencies for UK startups?

• What is the difference between a growth marketing agency and a fractional growth team?

• Who offers flexible growth marketing support without a retainer?

• Which UK marketing consultancies specialise in marketplaces?

• How much does startup growth marketing support cost?

Test them across the major platforms on a regular schedule.

Record:

• whether your brand appears

• whether it is recommended or simply mentioned

• which page is cited

• which competitors appear

• the position of the mention

• whether the information is accurate

• the sentiment

• the sources used

• changes since the previous test

Do not treat one isolated answer as a ranking.

Look for patterns across multiple related questions and repeated tests.

Measure citation share

A useful metric is your share of relevant cited answers.

For a fixed prompt set:

Citation share = answers citing your domain ÷ total answers tested

You can also track:

• brand mention share

• recommendation share

• accurate-answer share

• competitor citation share

• number of unique cited pages

• AI referral conversion rate

The goal is not simply to be mentioned more often.

It is to be mentioned accurately, positively and for commercially relevant questions.

Common AI optimisation myths

“SEO is dead”

It is not.

Search indexes, crawling, page quality, internal linking, relevance and authority remain central to how many AI systems retrieve information.

AI optimisation broadens SEO. It does not eliminate it.

“We need an llms.txt file”

An llms.txt file is a proposed convention for making website information easier for some AI tools to locate.

It may be useful in particular technical contexts, and maintaining one is unlikely to be harmful.

However, Google explicitly says it does not use llms.txt for Search or its generative Search features. Creating one will neither improve nor damage Google visibility.

Do not prioritise it ahead of crawlability, content, indexing, evidence or authority.

“There is special AI schema”

There is currently no universal schema markup that guarantees inclusion in AI-generated answers.

Use accurate, relevant structured data to describe the content and entities already present on the page.

Do not invent ratings, authors, prices, reviews or business information.

“We need to split every paragraph into tiny chunks”

Clear structure helps readers and systems locate information.

But there is no requirement to reduce every page into unnaturally short fragments.

Google says there is no ideal page length and no requirement to “chunk” pages specifically for its generative Search features.

Use the length and structure necessary to answer the question properly.

“Adding an FAQ section guarantees AI citations”

FAQs can be useful when they answer genuine customer questions.

A collection of repetitive, keyword-stuffed questions adds little value.

Use FAQs to cover important exceptions, objections and practical details that do not fit naturally elsewhere.

“AI-generated content automatically performs well in AI search”

Using AI to assist with research, structure or editing is not inherently a problem.

Publishing large volumes of generic, unverified content is.

Google warns that scaled AI-generated pages produced without additional value may violate its spam policies.

A human expert should still be responsible for:

• accuracy

• originality

• judgement

• evidence

• examples

• tone

• recommendations

“Ranking first in Google guarantees a ChatGPT citation”

Traditional rankings can help because search visibility makes a page easier to discover.

But an AI system may select a different source because it contains:

• a clearer answer

• more recent information

• stronger evidence

• better extractable passages

• more independent authority

• information better matched to a fan-out query

AI citation and traditional ranking are related, but they are not identical.

“We can optimise this once and forget it”

AI search products change quickly.

Competitors publish new evidence.

Crawlers update.

Pages become outdated.

Measurement tools improve.

AI optimisation should become part of ongoing SEO, content, PR and brand management rather than a one-off technical project.

A practical 90-day AI optimisation plan

Weeks 1–2: Technical and indexing audit

Check:

• Google indexing

• Bing indexing

• robots.txt

• noindex tags

• canonical tags

• XML sitemaps

• JavaScript rendering

• firewall rules

• AI crawler access

• internal links

• page speed

• mobile experience

• accessibility

Fix any issue preventing important pages from being discovered or understood.

Weeks 2–3: Entity audit

Review your:

• business name

• descriptions

• contact details

• About page

• team profiles

• social profiles

• Google Business Profile

• directory listings

• structured data

• review profiles

Resolve inconsistencies.

Weeks 3–4: Prompt and topic research

Create a list of:

• commercial questions

• comparison questions

• recommendation prompts

• customer objections

• implementation questions

• common misconceptions

• related fan-out questions

Test the most important prompts across several AI platforms and record the brands and sources currently appearing.

Weeks 4–8: Improve priority content

Start with pages that have the strongest commercial relevance.

For each one:

• answer the question clearly

• remove generic copy

• add original experience

• include evidence

• cover important comparisons

• explain limitations

• improve headings

• add named authors

• update sources

• strengthen internal links

• add accurate structured data

• connect the page to relevant services and case studies

Weeks 6–10: Create citation-worthy assets

Develop at least one asset competitors cannot easily reproduce.

For example:

• a benchmark report

• a customer survey

• a data-led case study

• a calculator

• an industry directory

• a practical template

• a detailed comparison

• a proprietary framework

Weeks 8–12: Earn external validation

Promote the new evidence through:

• journalists

• industry publications

• partners

• associations

• podcasts

• newsletters

• communities

• customer case studies

• relevant directories

Focus on relevance and credibility, not the total number of mentions.

Week 12: Measure and reprioritise

Review:

• AI citations

• brand mentions

• cited landing pages

• crawler activity

• referral traffic

• conversions

• Google AI impressions where available

• competitor visibility

• inaccurate information

Use the results to plan the next quarter.

AI optimisation checklist

Technical

• Important pages are indexed in Google.

• Important pages are indexed in Bing.

• AI search crawlers are not accidentally blocked.

• Firewalls and security tools permit verified crawlers.

• Pages return the correct status code.

• Canonical tags are correct.

• XML sitemaps are current.

• Important information is available in visible HTML.

• Mobile usability is strong.

• Forms, buttons and menus are accessible.

Content

• Each page answers a clear question.

• The answer appears early.

• The content adds something competitors do not.

• Material claims are supported with evidence.

• Statistics include context and sources.

• Comparisons use transparent criteria.

• Limitations are explained.

• Authors are identified.

• Important pages are reviewed regularly.

• Generic AI-generated filler has been removed.

Entity and brand

• The business name is consistent.

• The About page clearly explains the company.

• Team and expert profiles are complete.

• Contact and location details are accurate.

• Organisation structured data is present.

• Relevant external profiles are connected.

• Old or conflicting listings have been corrected.

Authority

• The business appears on relevant independent websites.

• Customers have left detailed, genuine reviews.

• Experts contribute to industry discussions.

• Case studies demonstrate real results.

• Original research or evidence is available.

• PR activity connects the brand to its areas of expertise.

• Important claims can be corroborated externally.

Measurement

• AI referral traffic is tracked.

• Conversions from AI traffic are measured.

• A fixed prompt set is tested regularly.

• Brand mentions and citations are recorded.

• Competitor visibility is monitored.

• Server logs are checked for crawler access.

• Inaccurate AI descriptions are identified and corrected at source.

The TL;DR

The businesses most likely to succeed in AI search will not be those that discover the cleverest markup trick.

They will be the businesses that become the clearest and most credible source of information in their market.

That means:

• making your website technically accessible

• answering real customer questions

• publishing information worth citing

• proving claims with evidence

• building a consistent business identity

• earning independent recognition

• keeping information current

• measuring visibility across multiple AI systems

Traditional SEO asked:

How do we rank this page?

AI optimisation adds a broader question:

Why should an AI system trust, use and cite us when answering this customer’s question?

Build your website and wider online presence around that question, and you improve more than your chances of appearing in ChatGPT or Google AI Overviews.

You create a more useful, credible and persuasive business presence everywhere your customers research you.