How to Get Cited by ChatGPT, Perplexity and Google AI Mode

What answer engines read, why they cite some sites and skip others, and the practical levers a B2B company can pull to show up in their answers.

By Godefroy de Compreignac , Founder of Bibop

October 5, 2026 7 min read

Key takeaways

  • An AI engine cites pages it can find, understand and trust. Every lever in this guide works on one of those three.
  • The pages that get cited answer a specific question in their first lines, with facts a reader can check.
  • Your robots.txt has to let answer engine crawlers in, or your pages stay out of reach.
  • The most reliable measurement is asking the same AI engines the same questions on a schedule and noting who gets cited.

A buyer asks ChatGPT “what’s the best payroll software for a 40-person company?” and gets a written answer with a handful of names and links. The companies named make the shortlist. Everyone else is invisible, at least for now.

More and more B2B buyers work this way, using ChatGPT, Perplexity or Google’s AI Mode to map a market before they ever talk to sales. The reasons an answer engine cites one site over another are fairly easy to read, and most of the levers are within reach of a small team. This guide walks through them.

How an answer engine builds its answer

When a question calls for current or specific information, answer engines all work in roughly the same way. They turn the question into one or more searches, query a web index (their own or a partner’s), read a few of the pages that come back, then write a summary and cite the sources it drew on.

That leads to three requirements.

  1. You have to be found. A page missing from the index, or blocked to the engine’s crawlers, never gets read. Google’s AI Mode draws on the Google Search index. ChatGPT and Perplexity run their own crawlers and rely on search indexes.
  2. You have to be understood. The model reads the text on the page. An answer buried in a long intro, or rendered only through JavaScript, is less likely to make the cut.
  3. You have to be trusted. Given two pages that say the same thing, the engine leans toward the source the rest of the web already recognizes and cites.

A model can also answer from memory, based on what it learned during training. In that case, what matters is how your company is described across the web, on your site and elsewhere. The levers below cover both cases.

Write pages that answer real questions

Start with the questions your customers actually ask. Your sales team hears them every week, and so does support: “does it integrate with our ERP?”, “how long does rollout take?”, “how is it different from that competitor?” Those are exactly the questions your prospects type into an AI tool.

For each one, a clear page does the job:

  • Lead with the answer. The first two or three sentences answer the question, and the rest of the page goes deeper. An engine looking for a specific answer finds it right away.
  • One heading per idea. Subheadings that mirror the follow-up questions (“How much does it cost”, “Who is it for”) help the model pull out the right passage.
  • Facts people can check. Pricing, timelines, limits, integrations, certifications. An AI engine is far more likely to quote a precise sentence than “an innovative, high-performance solution.”
  • One page per topic. A page that answers one question well beats a catch-all page that skims ten.

Product pages and industry pages matter as much as blog posts. A page titled “Scheduling software for clinics” that states clearly who it’s for, what the tool does and what it costs is an ideal source for a question like “what scheduling tool should a clinic use?”

A quick test

Read only the first three lines of your page. If they answer the question the page targets, an answer engine can use it.

Publish honest comparison pages

A large share of B2B questions are about choosing: “X or Y?”, “alternatives to X”, “best tool for…”. Answer engines like pages that put the options side by side, and they cite them often.

If you don’t publish a comparison, the engine will read someone else’s, possibly a competitor’s. A good comparison page:

  • covers several products, yours included, against explicit criteria (price, target company size, integrations, support);
  • says when another product is the better fit, which makes it credible to readers and models alike;
  • stays current, since an outdated price or a discontinued feature costs you the reader’s trust.

“Alternative to [competitor]” pages follow the same logic. They answer a real question asked by buyers who are already in the market.

Make your site readable by machines

Great writing stays invisible if crawlers can’t reach it or make sense of it. A few technical checks make a real difference.

Let AI crawlers in

Each engine sends its own crawlers, and your site’s robots.txt file decides which ones may read your pages. Plenty of sites block them without realizing it, through an overly broad rule or a default setting at their host or firewall. Here is an example that allows them explicitly:

# Answer engines
User-agent: OAI-SearchBot
Allow: /

User-agent: ChatGPT-User
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

# Google (classic search and AI Mode)
User-agent: Googlebot
Allow: /

Sitemap: https://www.example.com/sitemap.xml

Some vendors separate the crawler that collects training data (OpenAI’s GPTBot, for instance) from the one that reads pages to answer questions live. You can opt out of training and stay visible in search. On Google’s side, AI Mode relies on the Google Search index, so a page Googlebot indexes well is eligible. Also check your firewall or CDN, which can block these crawlers even when robots.txt allows them.

Serve your content as HTML

Many AI crawlers read the HTML your server sends back and don’t run JavaScript. If your text only appears after a script loads, they may see an empty page. A server-rendered or static site avoids the issue.

Add structured data

Structured data (Schema.org vocabulary, in JSON-LD) describes your content in a format machines read without ambiguity: your organization, your products and their prices, your articles with their authors and dates. It guarantees no citation, but it removes doubt about who you are and what you sell.

Get the basics right

Descriptive page titles and meta descriptions, an up-to-date sitemap, clean redirects and no broken links. These have served Google for years, and answer engines benefit too, since they rely on search indexes.

Check before you ship

A stray Disallow: / left over from a staging site is enough to hide an entire website. Reread your robots.txt after every redesign or migration.

Get mentioned on other sites

Your site says what you think of yourself. Answer engines also look at what others say about you. A company mentioned in reputable directories, independent comparisons, trade press or industry sites gains credibility, and it’s more likely to turn up among the sources the engine reads.

Some practical places to start:

  • Directories and marketplaces in your space. Software review platforms, your partners’ integration catalogs, recognized industry directories. Pick them for their reputation, not their number.
  • Sites already covering your topic. Niche blogs, newsletters, trade associations. A useful contribution or a customer story there is often worth more than a dozen anonymous links.
  • Consistent information. The same name, the same description, the same pricing everywhere. Conflicting details from one site to the next blur the picture the model forms of you.

This work pays off over time, in Google’s results and in what models remember.

Keep your pages fresh

Answer engines often favor recent information, especially on topics that move: pricing, features, regulation, tool roundups. A 2023 page still listing an old price loses out to a newer one.

A simple monthly routine is enough. Spot the pages losing rankings or traffic, update their numbers and examples, add the new questions customers are asking, then show the update date on the page and in your structured data. A real content update is what counts; changing the date alone does nothing.

Track your citations on a fixed set of questions

Standard analytics tools say very little about your presence in AI answers. Some visits do arrive with ChatGPT or Perplexity as the referrer, but many buyers read the answer without clicking, then come back later by searching your name.

The most reliable method is also the simplest:

  1. Pick a set of questions. Around thirty questions your customers really ask, from broad (“best software for…”) to specific (“X or Y for a 20-person team”).
  2. Ask the same engines on a schedule. ChatGPT, Perplexity and Google’s AI Mode, at a fixed interval, weekly for example.
  3. Record who gets cited. Your site, your competitors, and the third-party pages used as sources.
  4. Connect changes to your actions. A new comparison page, a robots.txt fix, a directory listing: watch what moves in the weeks that follow.

Answers vary from one run to the next, and that’s expected. The trend across several weeks and the whole question set is what matters. The sources cited for your competitors are a great source of ideas too: they show what kind of page the engine looks for on each question.

What Bibop does

Bibop runs this playbook every day for B2B startups and small businesses. Your agent tracks 50 of your customers’ questions in ChatGPT, Perplexity and Google’s AI Mode, writes the pages and comparisons you’re missing, fixes your site (titles, structured data, broken links, AI crawler access), keeps your pages current, gets you listed where your customers look, and measures the effect of every action. See the features page for the details, the technical page for the setup and the pricing page for plans.

Discover Bibop.

Private beta

Join the waiting list

Bibop opens new spots as it goes. Leave your address to receive your invitation.

You can unsubscribe at any time.

We use cookies to measure how our campaigns perform. Privacy policy