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.txt

• CDN 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:

1. Crawlability

2. Indexing

3. Clear website structure

4. Useful content

5. Internal linking

6. Original information

7. Author and company clarity

8. 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.