24/06/2026
If you run a startup, there is a new version of an old question:
“How do I get found?”
It used to mean Google.
Increasingly, it also means ChatGPT, Perplexity, Gemini, Google AI Overviews, AI Mode and other systems that can answer a question without sending someone through a traditional list of search results.
The temptation is to look for a new optimisation trick.
There isn't one.
OpenAI's own guidance is refreshingly straightforward: public websites can appear in ChatGPT Search, and if you want your content to be discoverable, surfaced and cited, you should make sure OAI-SearchBot can crawl your site. OpenAI also makes clear that search placement is not guaranteed.
Everything beyond that comes back to fundamentals:
Be useful. Be specific. Be clear about who you are. Make your claims verifiable. Build a body of information that other people can trust. And make sure search systems can actually access it.
That is less exciting than a secret GEO hack.
It is also much more likely to survive the next model update.
First, what does “getting cited by ChatGPT” actually mean?
There are two different things people often mix together.
ChatGPT can answer from what it already knows
Large language models have learned from enormous amounts of information. A model may know about a company, product or category without performing a live search in that conversation.
That does not necessarily mean your website will be cited.
ChatGPT can search the web
When ChatGPT Search is used, it can retrieve current web information and provide citations and links to sources. OpenAI says its search system ranks results using multiple factors intended to surface relevant and reliable information. There is no guaranteed position or formula that publishers can optimise for.
That distinction matters.
You cannot simply “rank #1 in ChatGPT.”
There is no equivalent of a single blue-link position that you can monitor.
Instead, you are trying to become one of the sources that an AI system considers useful when answering questions about your category.
And sometimes that source will be your own website.
Sometimes it will be a publication that mentions you.
Sometimes it will be a review, community discussion, interview, directory or another independent source.
That is why AI visibility is ultimately bigger than SEO.
Do I need “GEO” to get cited?
You will hear a lot about GEO, or Generative Engine Optimization, and the term is useful as shorthand for thinking about visibility in AI-generated answers.
But I would be careful with it.
It can make a fairly ordinary discipline sound like an entirely new branch of marketing.
Google's latest guidance makes the point clearly: the fundamentals of SEO continue to apply to AI Overviews and AI Mode, and there are no additional technical requirements or special schema markup needed to appear in those experiences. Google also explicitly says that you do not need to rewrite content into some special AI-friendly format.
That is important because it changes how I would approach this.
Do not create one website for Google and another version for AI.
Build a website that is:
Useful to people
Easy to crawl
Easy to understand
Specific about what it does
Supported by evidence
Clear about who created the information
Consistent across the wider web
Those qualities help everywhere.
So how does a startup actually become citable?
I would think about it in five layers.
1. Make yourself discoverable
If a search system cannot access your website, nothing else matters.
2. Make your company easy to understand
A search system should not have to infer what your company does from vague marketing language.
3. Publish information worth citing
Generic content gives an AI system very little reason to prefer you over thousands of other pages.
4. Build evidence outside your own website
Independent references help establish that your company, expertise and claims exist beyond your own marketing.
5. Keep the information accurate
A source that was useful six months ago can become a poor source if the product, pricing, market or underlying facts have changed.
None of these is a hack.
Together, they create something much more valuable:
a web presence that is easy to understand and difficult to dismiss.
1. Make sure ChatGPT can actually crawl your website
Start with the least glamorous part.
Your website needs to be accessible.
OpenAI's current publisher guidance specifically identifies OAI-SearchBot as the crawler used to discover content for ChatGPT Search. If you block it, you can prevent your site from being included in ChatGPT Search results.
That means checking:
robots.txtCDN and hosting rules
Security systems that may block crawlers
Important pages returning successful HTTP responses
Pages being accessible without authentication
Important content being present in crawlable page content
You do not need to blindly allow every AI crawler.
OpenAI distinguishes OAI-SearchBot from other crawlers, and Google's systems have their own controls. Your decisions should reflect what you actually want your content to be used for.
And this is where an important distinction is often missed:
OAI-SearchBot is about search discovery. GPTBot is associated with model training.
They are not interchangeable.
If your objective is specifically to appear in ChatGPT Search, the relevant OpenAI documentation is about OAI-SearchBot.
2. Get your website indexed properly
Being crawlable is not the same as being indexed.
For Google, that means making sure important pages can be discovered and indexed through normal Search fundamentals.
For Bing, Bing Webmaster Tools gives you visibility into crawling, indexing and search performance.
And there is now another reason to pay attention to Bing.
In 2026, Microsoft introduced AI Performance in Bing Webmaster Tools. It shows which pages from your site are cited in AI-generated answers across Microsoft Copilot, Bing's AI-generated summaries and selected partner experiences. It also exposes the grounding queries associated with that citation activity.
That is genuinely useful.
Not because Bing is a proxy for every AI engine.
It isn't.
But because it gives website owners something that has been missing for a long time:
first-party visibility into how their content is being used in AI answers.
You can see:
Which URLs are being cited
How citation activity changes over time
Which grounding-query themes are associated with your pages
Which pages receive the most citation activity
Bing is explicit that these metrics are citation measures, not rankings or authority scores. They also represent aggregated activity rather than a complete log of every AI answer.
That distinction is worth keeping.
3. Stop writing about your company like a brochure
This is probably the biggest content problem I see.
A startup website says:
“We are a next-generation AI-powered platform transforming the future of enterprise operations.”
Fine.
But if someone asks:
“What does this company actually do?”
the page does not give them a clean answer.
That is a problem for people.
It is also a problem for machines.
Your website should make basic facts extremely easy to verify.
For example:
What does the company do?
A direct answer.
Who is it for?
A specific audience.
What problem does it solve?
A concrete problem.
How does it work?
Enough detail to understand the mechanism.
Where is the company based?
If relevant.
Who founded it?
With a real profile.
What does it cost?
If pricing is public.
What integrations does it support?
If that matters to the buying decision.
Who uses it?
Where you can disclose this accurately.
You are not writing these pages for a language model.
You are removing ambiguity for everyone.
4. Create pages that answer real questions
A startup website often revolves around what the company wants to say.
A strong search strategy starts with what the customer wants to know.
Those are not always the same thing.
Suppose you sell compliance software.
Your website might be organised around:
Our platform
Our features
Our technology
About us
Your buyer may actually be searching for:
How do I automate SOC 2 evidence collection?
What does SOC 2 compliance software actually automate?
How much does compliance automation cost?
What is the difference between compliance automation and a compliance consultant?
How long does SOC 2 preparation take?
Which SOC 2 tools integrate with Jira?
Those questions are much closer to the actual information need.
And they create much better content.
This is also where our approach to ranking on Google without case studies becomes relevant.
You do not need to wait until you have dozens of customer logos before publishing useful information.
You need genuine knowledge of the problem you are discussing.
5. Give every important page a clear answer
This does not mean turning every article into a collection of tiny paragraphs designed for an AI crawler.
Google explicitly says there is no ideal page length and no requirement to break content into tiny chunks for AI systems. It recommends creating content for the audience, not for generative AI.
The principle is simpler:
Do not bury the answer.
If the question is:
What is the best social media platform for B2B startups?
Give the answer early.
Then explain the reasoning.
If the question is:
How much does X cost?
State the current price if you can verify it.
Then explain what affects the cost.
If the question is:
Does X integrate with Salesforce?
Say yes or no.
Then explain the integration.
Clear writing is not an AI trick.
It is good writing.
6. Use evidence instead of adjectives
This is where many startup websites lose me.
“Industry-leading.”
“Best-in-class.”
“Revolutionary.”
“Powerful.”
“Seamless.”
“World-class.”
These words are cheap.
Almost every company can say them.
What is harder is saying:
“The platform processes 2.4 million transactions per month.”
Or:
“We reduced the workflow from 14 manual steps to four.”
Or:
“The API supports X, Y and Z.”
Or:
“Our dataset covers 18,000 companies across six markets.”
Or simply:
“We launched in 2024 and currently serve customers in India, the UK and the US.”
The point is not to manufacture impressive numbers.
Use real numbers when you have them.
If you do not, write clearly without pretending you have evidence that you do not have.
That is more credible.
And if the information comes from somewhere else, cite the source.
7. Show who is behind the information
An anonymous blog is harder to trust than a clearly attributed one.
That does not mean every article needs an impressive list of credentials.
It means the reader should be able to answer:
Who wrote this?
And:
Why should I believe them?
For a founder-led company, this can be a real advantage.
Put the author's name on substantive articles.
Give the author a meaningful bio.
Link to their professional profile where appropriate.
Explain their relationship to the subject.
If an article is written by someone with first-hand experience of the problem, say so.
This is part of what Google describes through the broader concept of experience, expertise, authoritativeness and trustworthiness. But it is worth remembering that E-E-A-T is not a secret ranking score you can “optimise.” It is a useful way to think about whether your content deserves to be trusted. Google's current AI-search guidance continues to emphasise helpful, reliable, people-first content.
For La Rouge, this is particularly relevant.
Our content strategy work is built around the idea that founder expertise should become visible rather than hiding behind generic company copy.
8. Build a body of content, not isolated SEO articles
One article rarely establishes expertise.
A collection of useful pages around the same subject can.
Imagine a startup selling procurement software.
One article about procurement automation is fine.
A stronger body of work might cover:
What procurement automation actually means
Procurement automation vs traditional procurement
How to automate purchase approvals
Procurement software pricing
Procurement workflows for startups
How procurement integrations work
Common procurement automation mistakes
Procurement metrics worth tracking
A detailed comparison of competing approaches
Now the website is not simply targeting eight keywords.
It is becoming a useful source about a subject.
That distinction matters.
Google's latest generative-AI guidance explicitly encourages unique, valuable content and says existing SEO fundamentals remain foundational to visibility in AI search.
And it is consistent with the broader principle behind our search presence approach: visibility should be built as a system, not as a collection of disconnected optimisation tricks.
9. Make your internal links make sense
This is one of the easiest improvements a startup can make.
If you publish an article about founder-led marketing, and you already have a strong article about B2B social strategy, connect them.
If you discuss Google Search Console, link to your deeper explanation of Google Search Console platform properties.
If you discuss founder content, connect it to your guide on writing useful content when you have zero time.
If you discuss AI visibility, connect the topic to your broader search presence service.
Internal links do several useful things at once:
Help readers discover related information
Help search engines discover pages
Establish relationships between topics
Give important pages more contextual support
Make the site easier to navigate
Google explicitly recommends making content easily findable through internal links when discussing visibility in AI features.
This is not about adding links for the sake of adding links.
It is about building a coherent information architecture.
10. Earn information about your company outside your own website
This is where the conversation gets more interesting.
Your website can tell the world that your company is excellent.
That is expected.
An independent article saying your company is worth considering is different.
A founder appearing on a relevant podcast is different.
A customer discussing the product in a community is different.
A credible publication interviewing the founder is different.
A review on a relevant platform is different.
These references create context that your own website cannot create by itself.
But there is an important line here.
Do not manufacture mentions.
Google's current guidance explicitly warns against seeking inauthentic mentions across the web simply to influence generative AI visibility.
So I would not build a “100 mentions in 90 days” campaign.
Instead:
Contribute genuinely useful answers in relevant communities
Publish original research
Speak on relevant podcasts
Give interviews
Contribute expert commentary
Build relationships with journalists
Encourage honest customer reviews
Participate in industry conversations
Create something genuinely worth referencing
The objective is not to create artificial consensus.
It is to become actually known.
11. Publish original information when you can
This is one of the strongest opportunities for smaller companies.
Most companies rewrite the same information.
Someone publishes:
“10 trends in AI marketing.”
Another company rewrites it.
Then another.
By the twentieth version, nobody has added anything.
Original information is different.
You might have:
Customer research
Survey results
Internal benchmarks
Product usage patterns
Industry observations
Original experiments
Pricing data
First-hand lessons
A new framework you genuinely developed
Publish it.
Even a small dataset can be valuable if the methodology is clear and the conclusion is useful.
You do not need to pretend that every internal observation is a groundbreaking study.
Just be precise about what you actually know.
12. Do not hide important information behind JavaScript
This point is more nuanced than the old article made it sound.
Google can process JavaScript, and its current documentation explicitly says JavaScript frameworks can work for Search as long as standard JavaScript SEO practices are followed.
So “AI cannot read JavaScript” is too simplistic.
The practical advice is:
Make important information reliably accessible in the rendered page and do not create unnecessary barriers to crawling.
Your core:
Product explanation
Article text
Pricing
Key specifications
Company information
Author information
Important navigation
should not depend on a fragile interaction that only works after several client-side events.
This is good web development.
It is also good search hygiene.
13. Structured data is useful, but it is not an AI cheat code
This is another area where the industry has created unnecessary mystique.
Structured data can help search engines understand what a page represents and can make pages eligible for certain search features.
But it is not a secret instruction telling ChatGPT:
“Please cite this website.”
Google's current documentation is explicit that there is no special schema.org markup required for AI Overviews or AI Mode. It also says structured data should accurately match visible page content.
So use appropriate structured data where it makes sense.
Do not add invented entities, fake reviews or hidden claims.
And do not spend three days implementing an “AI schema strategy” while your homepage still does not clearly explain what the company does.
14. What about llms.txt?
You will probably encounter this recommendation if you research AI visibility.
My view is: do not make it a priority for Google.
Google's current generative-AI guidance says that maintaining llms.txt or similar files will neither help nor hurt visibility in Google Search because Google Search ignores them.
Other AI products may choose to use such files differently.
But if your resources are limited, I would prioritise:
Crawlability
Indexing
Clear website structure
Useful content
Internal linking
Original information
Author and company clarity
Independent references
before spending time on experimental machine-readable conventions.
15. Keep important information current
This matters particularly for startups.
Your pricing changes.
Your integrations change.
Your product changes.
Your positioning changes.
Your market changes.
An article written in 2024 saying your product supports three integrations is not a strong source in 2026 if the product now supports twelve.
That does not mean adding:
“Updated September 2026”
to an unchanged article.
Actually update it.
Replace obsolete information.
Add new examples.
Correct old claims.
Remove things that are no longer true.
Google's guidance encourages keeping content accurate and useful, while Bing's AI Performance documentation also points to freshness and accuracy as useful areas to review when improving content cited in AI answers.
Freshness is a change in substance, not a new date stamp.
16. Measure AI visibility without pretending it is more precise than it is
This is where I would change the old article most substantially.
You cannot reliably say:
“Our ChatGPT ranking is #3.”
There is no universal ranking position.
Responses vary by:
User
Location
Query wording
Search availability
Model
Time
Sources available at the moment
The way the question is framed
So build a monitoring set instead.
Take 20 to 50 questions your actual buyers might ask.
For example:
What is the best [category] software for a 50-person company?
How does [category] software compare with doing this manually?
What should a startup look for when choosing [category]?
Which companies offer [category] in India?
What are alternatives to [competitor]?
Then periodically check how your company appears across relevant AI systems.
Record:
Were you mentioned?
Were you cited?
Which page was cited?
How were you described?
Which competitors appeared?
Was anything inaccurate?
What source appeared instead of you?
This is directionally useful, not scientific measurement.
And where available, use first-party reporting alongside it.
Bing's AI Performance report now provides actual citation activity and grounding-query data for supported Microsoft AI experiences. It does not measure ChatGPT citations, so do not treat it as a universal AI visibility dashboard.
What I would not do
There is a growing industry around AI visibility, and some of the advice is getting unnecessarily elaborate.
I would not:
Stuff pages with keywords
Write naturally about the subject.
Add dozens of FAQs just because AI likes questions
Add FAQs when they genuinely help readers.
Rewrite every paragraph into two sentences
There is no evidence that an arbitrary paragraph length is an AI ranking factor.
Create fake reviews
Do not.
Buy meaningless backlinks
Do not.
Manufacture Reddit discussions
Do not.
Publish 50 thin AI articles
You are creating a larger version of the same problem.
Add llms.txt and assume the job is done
It is not.
Obsess over “GEO hacks”
If a tactic depends on a model behaving exactly the same way tomorrow, it is probably not much of a strategy.
Google's latest guidance is almost anti-hack in its framing: create useful, original content, maintain strong technical foundations, and apply the same core search principles that have always mattered.
The real opportunity for startups
This is the part I find more interesting than the technical side.
Large companies have an enormous advantage in traditional visibility.
They have:
More backlinks
More branded searches
More press
More reviews
More customers
More content
More years of accumulated authority
A startup cannot manufacture ten years of history.
But it can become extremely useful on a narrow subject.
That is a different game.
Instead of trying to be the most authoritative website about “AI,” become one of the clearest sources about:
AI procurement for mid-market companies.
Or:
No-code product validation for non-technical founders.
Or:
Cross-border SEO for Indian SaaS companies.
Or whatever your actual expertise is.
Go narrow enough that you can say something meaningful.
Then keep building.
One excellent article.
Then another.
Then an original dataset.
Then a podcast interview.
Then a founder perspective.
Then a customer story.
Then a useful comparison.
Eventually, there is a body of evidence.
That is much harder to manufacture.
And much harder for a competitor to copy overnight.
A practical 90-day plan
If I were helping a startup improve its AI visibility from scratch, I would not begin with a giant GEO project.
I would do this.
Month 1: Fix the foundation
Technical
Check robots.txt
Confirm OAI-SearchBot is not unintentionally blocked
Check Google indexing
Set up Bing Webmaster Tools
Submit the sitemap
Check important URLs
Make sure core content is accessible
Review internal linking
Add appropriate structured data where relevant
Content
Identify the 10 questions your best prospects actually ask.
Then audit whether your website answers them.
Not whether it mentions the keywords.
Whether it answers the questions.
Month 2: Build useful depth
Create or improve:
One strong page for each major buyer question
One deeper guide around your core category
One original piece of research or first-hand analysis
Author pages and bios
Relevant internal links between related pages
Then update the existing pages rather than continuously adding new ones.
Your objective is not “more content.”
It is better coverage of the subject you want to own.
Month 3: Build external evidence
Look outside your website.
Where can the company legitimately contribute?
Industry publications
Podcasts
Communities
Expert interviews
Reviews
Partnerships
Original research
Founder commentary
At the same time, start tracking your important buyer questions across AI search.
You will learn something valuable even when the answer is not “La Rouge is cited.”
You may discover that AI systems misunderstand your category.
Or your company.
Or your product.
Or the problem you solve.
That is useful information.
The uncomfortable truth
Getting cited by ChatGPT is not really a content trick.
It is a reputation problem.
If an AI system has to answer:
“What are the best companies doing X?”
and your company has:
A vague website
Three generic blog posts
No clear founder
No independent mentions
No useful product information
No original research
No clear explanation of what you actually do
there is very little reason for it to choose you.
Changing the H2 from “Benefits” to “Why does X matter?” is not going to fix that.
On the other hand, if you have built a genuinely useful, well-organised body of information around your category, your chances improve because there is simply more evidence available to understand and reference.
That is the work.
The simplest way to think about it
Do not ask:
“How do I make ChatGPT cite me?”
Ask:
“If ChatGPT had to explain my category to a potential customer tomorrow, would the web contain enough accurate, useful information about my company for it to include me?”
If the answer is no, you have your strategy.
Make the answer better.
Frequently asked questions
Can I guarantee that ChatGPT will cite my startup?
No.
OpenAI says ChatGPT Search ranks results using multiple factors and that placement is not guaranteed. There is no public formula that allows a company to guarantee citation.
Do I need to optimise specifically for ChatGPT?
Not in the sense of creating a separate technical SEO system.
OpenAI has specific crawler requirements for ChatGPT Search, particularly around OAI-SearchBot. Beyond that, the strongest approach is to build useful, clear, trustworthy web content and maintain strong search fundamentals.
Does SEO still matter for AI search?
Yes.
Google explicitly says its AI Overviews and AI Mode are grounded in its existing Search systems and that the established SEO fundamentals remain relevant. Pages need to be indexed and eligible to appear in normal Search.
Does ranking on Google guarantee that ChatGPT will cite me?
No.
Search systems and AI systems have overlapping inputs, but they are not identical. A strong Google ranking can be useful evidence of discoverability, but it does not guarantee inclusion in a ChatGPT answer.
Should I let OAI-SearchBot crawl my website?
If your objective is to make your public website eligible for inclusion in ChatGPT Search, yes, unless you have a specific reason to restrict it.
OpenAI says blocking OAI-SearchBot can prevent content from being included in ChatGPT Search.
Is GPTBot the same as OAI-SearchBot?
No.
OpenAI distinguishes its crawlers by purpose. OAI-SearchBot is used for search discovery, while GPTBot relates to access for model training. If your question is specifically about appearing in ChatGPT Search, OAI-SearchBot is the relevant crawler to check.
Does adding FAQ schema help ChatGPT cite my website?
There is no evidence that adding FAQ schema is a special way to make ChatGPT cite a page.
Structured data can help search engines understand pages and can make them eligible for certain search features, but Google explicitly says there is no special schema required for AI Overviews or AI Mode.
Use structured data when it accurately describes visible content and has a legitimate search purpose.
Should I create an llms.txt file?
It can be experimented with for systems that choose to support it, but it should not be treated as an essential Google AI-search optimisation.
Google currently says it ignores llms.txt for Search and that creating one neither helps nor hurts Google Search visibility.
How can I tell whether AI systems are citing my content?
For ChatGPT, you can manually test a representative set of real buyer questions and record the results over time.
For Microsoft AI experiences, Bing Webmaster Tools now has an AI Performance report showing cited pages, citation activity and grounding-query information.
Treat these as different measurement systems rather than assuming one dashboard represents every AI engine.
How long does it take to get cited?
There is no reliable universal timeline.
Crawlability and indexing problems can be addressed relatively quickly. Building useful content, topical authority and independent recognition takes longer.
Anyone promising that your startup will be cited by ChatGPT within a fixed number of days is selling certainty they cannot provide.
So, how do you get your startup cited by ChatGPT?
You make your company worth citing.
Then you make it easy to discover.
Then you make the information easy to understand and verify.
That means:
Let the relevant search crawlers access your site
Keep your technical SEO foundations clean
Explain exactly what you do
Answer the questions your buyers actually ask
Publish original information and first-hand expertise
Use evidence instead of empty claims
Put real people behind your content
Connect related information with useful internal links
Earn legitimate mentions outside your own website
Keep important information current
Measure AI visibility without pretending the data is more precise than it is
There is no shortcut around the fundamentals.
And honestly, that is good news.
Because a shortcut can be copied.
A company that has spent years becoming genuinely useful is much harder to copy.
That is the kind of visibility worth building.
