Home / Guides / GEO for technology companies
Guide · GEOHow to get your technology company cited by ChatGPT, Claude and other AI assistants
Generative engine optimisation (GEO) is the work of making your pages easy for AI assistants to find, understand and cite when buyers ask them questions. It builds on good SEO, and no one can guarantee a mention.
Take the checklist to your team.
A printable checklist: crawler access, answer-first writing, structured data, third-party evidence and the monthly test.
- The crawler and robots.txt checks
- Markup and entity checklist
- A monthly test you can run in an hour
How AI assistants choose what to cite.
When a buyer asks an assistant such as ChatGPT or Claude to recommend a partner or explain a topic, the assistant draws on what it learned in training and, where search is switched on, on pages it retrieves from the web. Pages that state the answer clearly, are easy to crawl, and are backed by other sources are easier to use and cite.
So GEO and SEO share most of their work. The difference is the standard: a page has to answer a question in a form a model can lift, not just rank for a keyword.
Seven things we do.
Let the crawlers in
Allow the search and user agents of the main assistants in robots.txt, and check that key pages are not blocked or hidden behind scripts.
Answer first
Put a short, direct answer under each question heading, then the detail. Define terms in one sentence.
Structured data
Mark up the organisation, services, articles and FAQs, so the page says plainly what it is.
A consistent entity
Use the same name, description and facts everywhere: site, directories, partner pages and profiles.
Third-party evidence
Earn mentions on sites assistants trust: directories, partners, publications and review sites.
Original data
Publish results and numbers only you have, with the method stated. Specific facts are what get quoted.
Fresh and attributed
Show dates, authors and sources, and update key pages on a schedule.
Which AI crawlers to allow.
They vary by purpose, so decide per agent. This is a starting list.
| Assistant | Agents in robots.txt | What they do |
|---|---|---|
| OpenAI (ChatGPT) | OAI-SearchBot, ChatGPT-User, GPTBot | Search results, user-triggered visits, and training |
| Anthropic (Claude) | Claude-SearchBot, Claude-User, ClaudeBot | Search results, user-triggered visits, and training |
| Perplexity | PerplexityBot, Perplexity-User | Search index, and user-triggered visits |
| Google (Gemini) | Google-Extended | A control for use of content in Google's AI models |
How to tell whether it is working.
Keep a fixed list of 20 to 30 questions your buyers ask. Run them in each assistant every month and log whether your firm is mentioned, cited, or absent, and which competitors appear instead. Add referral traffic from the assistants in analytics. Expect movement to be slow and uneven, because answers differ from one run to the next.
Formats that assistants can lift.
Assistants build answers from passages. A passage gets used when it is self-contained, specific and easy to attribute. These formats do well because they package an answer cleanly.
- A definition in one sentence. Start a page or section with a plain statement of what something is.
- A direct answer under a question heading. Put the answer first, then the detail.
- A comparison table. Options, criteria and a verdict, with the rows labelled.
- A numbered process. Steps in order, each with a short explanation.
- A short FAQ. Real questions buyers ask, with answers of two or three sentences.
- Original data. Your own results and numbers, with how you measured them. Specific facts are quoted; vague claims are not.
Say plainly what each page is.
Structured data is code that labels your content for machines. For a services company, the useful types are Organization (who you are, with your logo, address and social profiles), Service (what you offer), Article (guides, with author and dates), FAQPage (questions and answers on the page) and BreadcrumbList (where the page sits). Use the sameAs property to link your official profiles, so systems can connect your site to your LinkedIn page and other listings.
Structured data does not guarantee a citation. It removes ambiguity, which makes your pages easier to use. Validate it with Google's Rich Results Test or the Schema.org validator, and make sure the markup matches the visible text on the page.
A starting robots.txt.
Allow the crawlers that serve buyers, and decide separately about those that collect training data. This example allows search and user-triggered agents. Check each vendor's current documentation, because agent names and behaviour change.
User-agent: OAI-SearchBot Allow: / User-agent: ChatGPT-User Allow: / User-agent: Claude-SearchBot Allow: / User-agent: Claude-User Allow: / User-agent: PerplexityBot Allow: /
Two points are often misunderstood. Google's AI Overviews and AI Mode draw on the same index as Google Search, so ordinary SEO and Googlebot access are what matter, and no special markup is required. And Google-Extended is a control for use of your content in Google's AI models; Google states it does not affect inclusion or ranking in Search. Also confirm that your CDN or firewall is not blocking these crawlers by mistake.
Be mentioned where assistants look.
Assistants weigh what other sources say about you. For a technology services firm, the useful places are directories and review sites (Clutch, G2 and similar), partner and marketplace listings for the platforms you work with, industry publications, conference and podcast pages, and a complete, consistent LinkedIn profile.
Keep the facts identical everywhere: company name, founding year, services, locations, leadership names and a one-line description. Contradictions across sources lower confidence in all of them. Ask clients and partners to describe you in their own words in public places, because independent mentions carry more weight than your own claims.
A monthly test you can run in an hour.
Create a list of 20 to 30 questions your buyers ask, in the form they would type: "best SAP migration partner for a mid-size manufacturer", "how do we measure lead quality in B2B?". Run them in each assistant monthly and log the result.
| Question | Assistant | Mentioned? | Cited? | Who appears instead | Action |
|---|---|---|---|---|---|
| Best SAP migration partner for manufacturers | ChatGPT | No | No | Two named competitors | Write an industry page with proof |
| What is lead-to-SAL ratio? | Claude | Yes | Yes | Two other agencies | Keep the page fresh |
Answers differ from one run to the next, so look at trends across the full list, not single results. Add referral traffic from the assistants in GA4, and ask new leads how they found you.
What GEO can and cannot do.
It cannot be guaranteed
Nobody controls what an assistant says. Treat claims of guaranteed placement as a warning sign.
It rewards good SEO
Most of the work, such as clear pages, structure, proof and links, helps ordinary search too.
It moves slowly
Retrieval is quick, but a change in how models describe you can take months and is uneven.
It carries risk if abused
Hidden text, keyword stuffing and fake reviews can harm you in both search and AI systems.
Where to start if you have done nothing yet.
Baseline and access
Write your list of 20 to 30 buyer questions and run it in each assistant. Check robots.txt and your firewall for blocked crawlers. Audit your top ten pages for a clear definition and a direct answer. Fix the technical blockers first.
Rewrite and mark up
Rewrite service pages and top articles answer-first. Add FAQs with real questions. Add Organization, Service, Article and FAQ markup. Align your company description across your site, LinkedIn and directories.
Evidence and review
Complete directory and partner profiles. Publish one piece of original data, such as an anonymised result or a survey. Re-run the question list and compare with the baseline. Choose the next ten questions to win.
Four page types assistants cite for services firms.
Definition pages
Plain explanations of terms your buyers ask about, such as lead-to-SAL ratio or SAP S/4HANA migration, each starting with a one-sentence definition.
Comparison and decision pages
Options laid out against criteria, for example build versus buy or two platforms, with a clear verdict for each type of buyer.
Process and cost pages
How a project runs, how long it takes and what drives the price. Buyers ask assistants these questions constantly.
Evidence pages
Case studies with numbers, method notes and dates. These are what assistants quote when asked who has done something before.
Take the checklist to your team.
A printable checklist: crawler access, answer-first writing, structured data, third-party evidence and the monthly test.
- The crawler and robots.txt checks
- Markup and entity checklist
- A monthly test you can run in an hour
Algoworks
See how assistants describe your firm today.
We will run your buyers' questions across the main assistants and show you where you appear, and where you do not.
- A 30-minute callYou describe your funnel and where it leaks.
- A look at your funnel dataRead-only access to analytics and your nurture platform.
- First findingsWhat we would change first, and what it should move.