How to Get Cited by ChatGPT and Perplexity: a Guide for Startups

More people now ask an AI assistant which tool to use before they search Google. When ChatGPT or Perplexity answers "what is a good invoicing tool for freelancers", it names a few products and links to its sources. If your startup is one of them, you get a visitor who already has a clear need. If it is not, the answer names your competitors instead.

Getting cited is not magic and it is not a separate discipline from SEO. Most of it is three jobs. Make sure assistants can reach your pages. Write those pages so a sentence can be lifted straight into an answer. And get mentioned on the pages assistants read when they research your category. The practice has a name, generative engine optimization (GEO). This guide covers the parts a startup can act on this week.

How assistants choose what to cite

There are two ways an assistant can know about you, and they work differently.

• Live search. When ChatGPT search or Perplexity answers a question, it runs searches, fetches a handful of pages, and writes the answer from what it read, with links to those pages. To be cited this way, your page has to be in the index the assistant searches, it has to be fetched, and it has to state the needed fact clearly.
• Training data. When an assistant answers from memory, it relies on what was in its training data. You cannot control this directly. Being described consistently across many pages on the web, over time, is what puts you there.

Live search is where a startup can make a difference quickly, so most of this guide is about it. The good news: pages that rank well in search engines tend to get retrieved by assistants too. The work overlaps.

Step 1: let the right crawlers in

Each AI company runs its own crawlers, and many sites block them by accident, through a hosting setting, a firewall rule or a robots.txt line copied from a template. Know which is which before you allow or block anything.

Company	For search answers	For training
OpenAI	OAI-SearchBot (ChatGPT search), ChatGPT-User (pages a user asks it to open)	GPTBot
Perplexity	PerplexityBot (its index), Perplexity-User (pages fetched for an answer)	Perplexity says PerplexityBot is not used for training
Anthropic	Claude-SearchBot, Claude-User	ClaudeBot
Google	Googlebot (AI Overviews use Google's search index)	Google-Extended controls Gemini training use

The key point: blocking a training crawler does not block the search crawler. You can block GPTBot and still appear in ChatGPT search, as long as OAI-SearchBot is allowed. If you block OAI-SearchBot or PerplexityBot, you are asking to be left out of those answers.

Check your setup with the robots.txt tester, which shows whether Googlebot, GPTBot, ClaudeBot and PerplexityBot may crawl a page. If you use a CDN or firewall with bot protection, check its AI crawler settings too: a block there overrides a robots.txt that says yes.

Step 2: make the page readable without JavaScript

Many AI crawlers fetch the raw HTML and do not run JavaScript. If your page builds its content in the browser, they see an empty shell. A product page that shows nothing until a script loads can be invisible to an assistant even when it looks fine to you.

• Server render or pre-render the main text of every page you want cited.
• Check what a crawler sees. View the page source (not the inspector) and look for your main text. Or paste the URL into the web page to Markdown converter, which shows the main content as a crawler would extract it.
• Avoid text inside images for anything you want quoted, such as pricing, features or the one-line description.

Step 3: write sentences an assistant can quote

An assistant writing an answer looks for a sentence that states the fact it needs, plainly, near the top of a section. Pages that bury the answer in marketing language get passed over for pages that just say it.

• Answer first. Under a heading that asks a question, the first sentence should answer it. "Acme costs $12 a month per user, with a free plan for one user" beats three paragraphs on value.
• Say what it is and who it is for in one sentence on the homepage: "Acme is invoicing software for freelance designers."
• Use specific facts. Prices, limits, supported platforms, integrations, founding year, location. Vague claims cannot be cited; facts can.
• Use tables for comparisons and lists for steps. Structured content is easier to extract correctly.
• Write the pages people ask about: pricing, comparisons with named alternatives, "how to do X with Acme", and an FAQ that uses the questions customers actually ask.
• Show dates. A visible "last updated" date helps an assistant judge whether a fact is current.

Add structured data as well: Organization on the homepage, SoftwareApplication or Product on product pages, Article on guides. It gives machines the same facts in a form that needs no interpretation.

Step 4: get into the indexes assistants search

An assistant can only fetch a page it can find. ChatGPT search draws partly on third-party search indexes, including Bing's, as well as OpenAI's own crawler. Perplexity runs its own index. Google's AI Overviews use Google's index.

• Get indexed in Google and Bing. Verify the site in both Search Console and Bing Webmaster Tools, and submit your sitemap to both. Our guide on getting indexed by Google covers the steps.
• Use IndexNow for Bing. It notifies Bing and other engines the moment a page changes. The IndexNow tester sets it up.
• Rank for the questions. When an assistant searches "best invoicing tool for freelancers", the pages it fetches are mostly the ones ranking for that search. Ordinary SEO is part of GEO.

Step 5: be mentioned where assistants look

For "which tool should I use" questions, assistants often cite pages that compare several products, not the products' own sites. Your own site says you are good. Other pages saying so is what gets you into the answer.

• Comparison and "best of" lists. Find the articles that rank for "best [your category]" and "[competitor] alternatives", and ask their authors to consider your product, with a short note on what is different.
• Directories and launch platforms. A listing on a crawled platform describes your product on a page assistants can fetch. Pick ones that fit from the launch platform database.
• Communities. Genuine discussions on Reddit, Hacker News and niche forums often appear in AI answers. Take part honestly; planted recommendations get noticed and removed.
• Consistent descriptions. Use the same one-line description of your product everywhere: your site, listings, profiles, press kit. Consistency helps assistants connect the mentions to one product and repeat the description correctly.

Step 6: add an llms.txt file (a small bet)

llms.txt is a proposed Markdown file at the root of your site that gives language models a short summary of the site and links to its most useful pages. It is cheap to add and does no harm.

Be honest with yourself about what it does. It is a proposal, not a standard, and support varies: no major assistant has publicly committed to using it to choose citations. Treat it as a tidy summary for any tool that reads it, not as a ranking lever. The llms.txt generator builds a valid file from your product name, summary and key pages in a few minutes, and the llms.txt validator checks an existing one.

Step 7: run the GEO checker

The GEO checker tests a page against the points above in one pass. It checks whether ChatGPT, Claude, Perplexity and Google's crawlers may fetch it, and whether the content is readable without scripts. It also looks for an llms.txt file, structured data and a clear page structure. Run it on your homepage, pricing page and main feature pages, and fix what it flags.

How to measure it

There is no Search Console for AI assistants yet, so measure from three sides.

• Referral traffic. In your analytics, look for visits from chatgpt.com, perplexity.ai, claude.ai and gemini.google.com. ChatGPT often adds utm_source=chatgpt.com to the links it shows, which makes those visits easy to separate.
• Server logs. Count requests from OAI-SearchBot, ChatGPT-User, PerplexityBot and Perplexity-User. A user-triggered fetch means someone's question led the assistant to your page.
• Manual checks. Write down ten questions a customer would ask an assistant, such as "best tool for X" or "how do I do Y". Ask them in ChatGPT, Perplexity and Claude once a month, and note who is cited. Answers vary between runs, so look at the pattern over months, not a single answer.

What does not work

• Hidden text aimed at AI. Instructions hidden in a page for language models are a form of hidden text, which search engines treat as spam. A page that gets dropped from the index cannot be retrieved at all.
• Mass-produced pages. Hundreds of thin AI-written pages are as easy for assistants to ignore as they are for search engines.
• Blocking all AI crawlers, then expecting citations. Decide what you allow on purpose, crawler by crawler.
• Fake reviews and planted forum posts. They break platform rules and, once found, damage the trust you are trying to build.

Where Aback Launch fits

Every product on Aback Launch gets a permanent launch page that search engines and AI assistants can read. The product's description, category and link are in plain, server-rendered HTML. Launches go live every Friday at 00:00 UTC. It is one more consistent description of your product on a page assistants can fetch.

FAQ

How do I get my website cited by ChatGPT?

Allow OAI-SearchBot in robots.txt and in your firewall, and make sure your main content is in the HTML without JavaScript. Write pages that state facts plainly near the top, and get indexed in Bing and Google. Then get your product mentioned on comparison pages and directories that rank for your category.

Does blocking GPTBot stop ChatGPT from citing my site?

No. GPTBot is OpenAI's training crawler. ChatGPT search uses OAI-SearchBot. You can block one and allow the other.

Does llms.txt help with AI citations?

It is unproven. llms.txt is a proposal, and no major assistant has publicly confirmed using it to pick sources. It is cheap to add and does no harm, so treat it as a small bet.

How long does it take to get cited by AI assistants?

For live search answers, as soon as your page is indexed where the assistant searches and ranks for the question. That can be weeks for a narrow question. Appearing in answers from training data takes much longer and is outside your direct control.

How do I know if ChatGPT or Perplexity sends me traffic?

Check your analytics for referrals from chatgpt.com and perplexity.ai, and look for utm_source=chatgpt.com. Your server logs also show visits from their crawlers and user-triggered fetchers.