How to Get Cited by ChatGPT, Perplexity and AI Overviews

Most startups that ChatGPT never mentions are not losing a quality contest. They are failing a fetch. The crawler that would have read the page got a 403 from a firewall rule, or it got an empty <div id="root"> because the content only appears after JavaScript runs, and a page the engine cannot read is a page it cannot cite.

Fix the fetch, and you are competing on the part that actually decides citations: whether several independent sources say the same thing about your product. Here is the order to work through it, with the checks that take five minutes each.

How AI answers pick their sources#

An answer engine can mention you in two ways, and only one of them responds to anything you do this quarter.

The first is training data: what the model absorbed before it shipped. Getting into that takes a training cycle, many months, and you cannot request it.

The second is retrieval. When someone asks ChatGPT, Perplexity or Google a question that needs current information, the engine runs searches, pulls a handful of pages, and writes an answer from the passages that agree with each other. ChatGPT search draws on OpenAI's own OAI-SearchBot crawl plus third-party search providers, historically led by Bing. Perplexity runs its own crawler and index. Google AI Overviews are built from Google's own index, which is why the same page that ranks tends to be the one that gets quoted.

The consequence is simple. A product named on five independent pages, each describing it the same way, beats a product described brilliantly on one page. Retrieval engines are looking for agreement, and your own site is only one vote.

Step 1: let the right crawlers in#

Each engine uses different user agents for different jobs, and the difference matters because blocking the wrong one removes you from live answers.

Robots.txt is only half of it. The other half is the bot protection on your CDN or host, which can reject these agents before robots.txt is ever read. Cloudflare added a one-click switch for blocking AI crawlers, and on some zones it is on by default, so check it rather than assuming.

Test the real response, not the config file:

curl -s -o /dev/null -w "%{http_code}\n" -A "OAI-SearchBot/1.0" https://yourdomain.com/pricing
curl -s -A "PerplexityBot/1.0" https://yourdomain.com/pricing | grep -c "your product name"

You want a 200 and a non-zero count. A 403, a challenge page, or a count of zero means that engine cannot see you.

Step 2: put the answer in the HTML#

Googlebot renders JavaScript. Most AI crawlers do not: they fetch the raw HTML and move on. If your marketing site is a client-rendered single-page app, the pricing table, the feature list and the FAQ may not exist in the version those crawlers read.

The check is "view source", not the browser's element inspector. Open view-source on your homepage and your most important feature page, search for a sentence from the body copy, and if it is not there, the page is invisible to retrieval. Static generation or server rendering fixes it. Every page on this site is statically generated for exactly this reason, so the HTML a crawler receives is the full article.

Step 3: write answers an engine can lift#

Retrieval engines quote passages, not pages. The passage they pick is usually the one that answers the question in a sentence or two without needing the paragraph before it.

So write the first two sentences under every H2 as if they will be pasted somewhere without context. Name the entity, state the specific fact, then explain.

Before: "There are a lot of ways to think about pricing, and it really depends on your team."

After: "Acme costs $12 per seat per month, with a free plan for up to three users. Teams over 50 seats get volume pricing on request."

The second version contains a product name, two numbers and a condition, and it survives being quoted on its own. Question-shaped headings help for the same reason: an H2 that reads "How much does Acme cost?" matches the question someone typed almost word for word. This is the same specificity that makes low-competition, question-shaped keywords winnable in Google at a low Domain Rating.

Step 4: add structure, and keep expectations honest#

FAQ blocks with FAQPage schema are the cheapest structural win. Google restricted FAQ rich results to a small set of authoritative sites back in 2023, so do not add them for the search snippet. Add them because an explicit question paired with a two to four sentence answer is the easiest possible thing to extract. Every post on this blog, including this one, ends with one for that reason.

Organization schema with sameAs links to your X, LinkedIn, GitHub, Crunchbase and Product Hunt profiles tells engines that those profiles are the same entity as your site. It costs one JSON-LD block.

Then there is llms.txt. "llms.txt" gets around 5,400 US searches a month, more than any other query in this space, and as of this month none of the major answer engines has confirmed it influences which sources they cite. Add it if you like. Do not do it first.

Step 5: build consensus off your own site#

This is where most of the result comes from, and it is the step founders skip because it does not happen in their codebase.

The sources that retrieval engines cite for "best X for Y" questions are overwhelmingly third-party: directories and review platforms, comparison posts, Reddit threads (Perplexity cites Reddit heavily), and YouTube videos with transcripts. You want to be named on as many of those as you can, with the same one-line description each time.

That last part is the lever. If G2 calls you "an invoicing tool for freelancers", Product Hunt calls you "AI billing for creators" and your homepage says "the financial OS for independents", no engine can confidently place you in a category. Pick one sentence with your category words in it, and use it on every listing.

The fastest way to get fifteen or twenty consistent third-party pages is a curated directory batch. The case for doing it early, and why platforms like G2 and Crunchbase double as consensus signals, is in the directory submissions playbook. The specific list and the order to submit in is the launch week checklist. The same links lift your Domain Rating, which is covered phase by phase in the DR roadmap.

Step 6: get indexed where the engines search#

An engine that retrieves from Google's index cannot cite a page Google has not indexed, and the same goes for Bing. Two habits cover both:

1. Google: submit your sitemap in Search Console and inspect every new URL on the day it ships. The weekly routine is in the five Search Console reports post.
2. Bing: verify the site in Bing Webmaster Tools, then submit new and changed URLs through IndexNow. It is a single POST request and a key file in your public folder, and it gets pages into the index that ChatGPT search leans on without waiting for a crawl.

How to measure AI citations#

Two numbers, checked monthly.

Referral traffic. In GA4, filter sessions by source for chatgpt.com, perplexity.ai, gemini.google.com and copilot.microsoft.com. ChatGPT also appends utm_source=chatgpt.com to the links it cites, so those visits are labelled even when the referrer is stripped.

A fixed prompt set. Write ten prompts your buyer would actually type ("best invoicing tool for freelance designers in the EU", not "Acme review"). Run each one in ChatGPT, Perplexity and Google once a month and record three things: whether you are named, which URL is cited, and which competitor is named instead. Keep the prompts fixed so month-over-month changes mean something.

A page that gets cited for a neighboring prompt but not the one you want is the AI version of a page-two ranking, and it responds to the same treatment: a targeted content refresh that adds the missing answer rather than a new page that starts from zero.

The order to do it in#

Steps one to four are an afternoon of work on your own site. Step five is the slow part, and it is the one our directory submission service exists to take off your plate. For the rest of the system, from keyword research to measurement, grab the free guide and follow new breakdowns on X.