How to Get Your Startup Cited in AI Search Answers (AEO Playbook)

Open ChatGPT and ask for the best tool in your category, the way a buyer would. You'll get a tidy list of five or so products. If yours isn't on it, that answer is quietly sending your buyers somewhere else. AI answers look authoritative, but they're built from whatever the assistant could find and trust at the moment someone asked.

The good news: this isn't a brand-budget game yet. A new product with almost no domain authority can get mentioned and linked if the assistant can reach its pages, understand them in one read, and find it named on sites it already uses. Here's the playbook I'd run on a brand-new product with no budget.

Quick answer

To get your startup cited in AI search answers, first make sure the search crawlers behind ChatGPT, Perplexity, Claude and Google can actually reach your site. Then rewrite your key pages so each one answers a real buyer question in plain text in the first paragraph, with one clear sentence saying what your product is and who it's for. Finally, earn honest mentions on the pages assistants already pull from: launch platforms, directories, helpful forum answers and other people's "best X for Y" lists. Measure it by running the same 10 to 20 buyer prompts every month and by watching AI referrals in your analytics.

How AI answers actually pick their sources

Most bad "AEO" advice skips this part. When an assistant answers with web search on, it roughly does three things:

1. Rewrites the question into searches. One messy question becomes a few short queries. Google describes its AI Overviews and AI Mode doing a "query fan-out" across subtopics, and OpenAI's help docs describe ChatGPT search rewriting prompts into targeted queries it sends to search providers, with Bing named among them.

2. Pulls candidate pages from an index. Crawlers build that index. If they couldn't fetch your page, you're not in the pool.

3. Reads a handful of pages and writes the answer. Citations go to the pages that supported specific claims. A page that states the fact plainly is easier to cite than one that hints at it across six sections.

When search isn't triggered, the model answers from training, which you can't change quickly. It's shaped by how consistently your product was described across the web over time.

So you have two levers: be retrievable (crawlable, indexed, quotable) and be mentioned on pages that get retrieved for your category.

Step 1: Check where you stand before you change anything

Start with a baseline, or you'll never know what worked.

Write 10 to 20 real buyer prompts

Use your buyers' words, pulled from sales calls, support emails and Reddit threads. Mix three types:

• Category prompts: "best invoicing tool for freelance designers", "simple time tracking app for small agencies"

• Problem prompts: "how do I stop clients paying invoices late as a freelancer"

• Comparison prompts: "FreshBooks alternative that's cheaper for one person", "[your product] vs [competitor]"

Add two or three prompts with your product's name in them. That's often where wrong pricing or an old tagline shows up.

Run them across three assistants and log the results

For most indie SaaS that's ChatGPT (with search), Perplexity and Google's AI Overviews or AI Mode. Use a fresh or logged-out session where you can, and log every run:

That last column is the gold. It shows which third-party pages the assistant trusts for your category, and those become your targets in Step 4. Answers vary between runs, so read the pattern, not one result.

Step 2: Make sure the right crawlers can reach you

This boring step quietly kills a lot of small sites. Each assistant uses different crawlers:

• ChatGPT search uses OAI-SearchBot. OpenAI's crawler documentation says sites that opt out of OAI-SearchBot won't be shown in ChatGPT search answers. GPTBot is a separate crawler for model training, and the two settings are independent.

• Perplexity uses PerplexityBot to surface and link sites in its results. Perplexity's crawler docs say it isn't used for training foundation models, and they include notes on allowing it through a WAF.

• Claude uses Claude-SearchBot for search indexing, separate from ClaudeBot (training) and Claude-User (fetches a page when a user asks). Anthropic's help center says blocking Claude-SearchBot may reduce your visibility in Claude's search results.

• Google AI Overviews and AI Mode use normal Googlebot and normal Search eligibility. People get this one wrong. Google-Extended is a separate token that controls whether Google can use your content for Gemini training and for grounding in the Gemini app and Vertex AI. Google says it does not affect your inclusion in Google Search, and Google's guide to AI features and your website says that to appear as a supporting link in AI Overviews or AI Mode, a page just has to be indexed and eligible to show a snippet. No extra requirements.

If your robots.txt has a blanket Disallow: / from an old template or plugin, fix that first. You don't have to name every bot. A plain User-agent: * group already lets them in. Naming them helps when you want to be explicit, but remember a crawler follows only the most specific group that matches it, so repeat your Disallow lines there:

# Search and answer crawlers: welcome
User-agent: Googlebot
User-agent: Bingbot
User-agent: OAI-SearchBot
User-agent: PerplexityBot
User-agent: Claude-SearchBot
Disallow: /app/
Disallow: /admin/

# Training crawlers are a separate choice.
# Blocking GPTBot does not remove you from ChatGPT search.
# Blocking Google-Extended does not remove you from Google Search,
# but it does opt you out of grounding in the Gemini app.
# User-agent: GPTBot
# Disallow: /

User-agent: *
Disallow: /app/
Disallow: /admin/

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

Two more checks people skip:

1. Your CDN or host. robots.txt can say "allow" while a firewall or "block AI bots" toggle says no. Google's guide lists allowing crawling "by any CDN or hosting infrastructure" as a basic. Check your logs for 403s to these user agents.

2. Bing. Since OpenAI names Bing as one of ChatGPT search's providers, set up Bing Webmaster Tools and submit your sitemap. It takes ten minutes and is easy to skip at launch.

Step 3: Make your pages easy to quote

Assistants cite pages that state the thing plainly. Most startup homepages say "the future of work, reimagined" and make you scroll to learn it's a scheduling app.

Write one definition sentence and use it everywhere

The format: [Product] is a [category] for [specific buyer] that [main job, ideally with a concrete detail].

PaperTrail is an invoicing tool for freelance designers that turns a finished project into a client-ready invoice and chases late payments for you.

(PaperTrail is a made-up example.) Put your version on your homepage, about page, meta description, launch listings and bios. Same words, same name, same spelling. "PaperTrail" here and "Paper Trail App" there makes the assistant guess whether they're the same thing. If the sentence won't come, the problem is usually positioning, not copy. A one-page positioning brief sorts that out in an afternoon.

Put the answer first on every page that matters

For each key page, ask which question it answers, then answer it in the first two or three sentences. Pricing page: what it costs and what's included. Use-case page: who it's for and what changes. Docs: lead with the steps.

Keep facts in plain HTML, with dates

• Pricing, plan limits and integrations go in real text, not inside a screenshot or a JavaScript widget that only renders after a click.

• Comparison facts go in an HTML table. I covered the format in how to write a SaaS comparison page that ranks without sounding like a brochure.

• Date anything that changes: "Pricing as of October 2026". It stops old numbers getting quoted forever.

Add an FAQ that answers questions people really ask

Take questions from your support inbox, not the ones you wish people asked. "Does it work with Stripe?" "What happens to my data if I cancel?" One question, one direct answer of two to four sentences. That's an easy shape for an assistant to lift.

Step 4: Earn mentions where assistants already look

Go back to the "source pages it used" column. That's your list.

1. Honest listings on launch platforms and directories. Start with ones from your sheet. Use your definition sentence, real screenshots and current pricing. Ten accurate listings beat fifty copy-pasted ones.

2. Helpful forum and Reddit answers. If threads show up in your source column, find ones asking about the exact problem you solve and answer properly, product or not. When it's genuinely relevant, say you built it. My notes on what still works on Reddit for indie SaaS without getting banned cover how to do this without getting flagged as spam.

3. Other people's "best X for Y" lists. If the same blog lists keep getting cited, email their authors: what it does, who it's for, one honest reason it belongs, and a free account to test. One yes can beat a month of blog posts.

4. Curated GitHub lists. For dev tools, "awesome" lists and curated repos are read by people who care. Follow each list's contribution rules.

Step 5: Add structured data, but keep it honest

Schema helps machines read your page. It's not a citation button. Google's guide says no special schema.org markup is needed for AI Overviews or AI Mode, and structured data should match the visible text. Worth adding:

• Organization on the homepage, with name, logo, URL and the profiles you use (sameAs).

• SoftwareApplication on your product or pricing page, with category, operating system and real price.

• FAQPage only where there's a visible FAQ. Google limited FAQ rich results to well-known government and health sites in 2023, so don't expect dropdowns in results.

Don't mark up reviews you don't have, ratings you invented or features you haven't shipped.

What to measure

• AI referrals in analytics. OpenAI's publisher FAQ says ChatGPT adds utm_source=chatgpt.com to referral links from ChatGPT search. In GA4, go to Traffic acquisition, switch to Session source, and search for chatgpt.com. Check referrals for perplexity.ai and other assistant domains too. Some AI clicks arrive without a referrer, so treat these numbers as a floor.

• Search Console. Google counts AI Overviews and AI Mode traffic inside the normal "Web" search type in the Performance report, so watch the pages you rewrote.

• Your prompt sheet, monthly. Same prompts, same assistants. Count mentions, links and new source pages. This is your real scoreboard.

• One signup question. Add "ChatGPT or another AI assistant" to "How did you hear about us?"

A 30-day sprint for a solo founder

Week 1: baseline and access

• Write 15 buyer prompts and run them on three assistants. Fill the sheet.

• Check robots.txt, CDN bot settings and server logs for the crawlers above.

• Verify the site in Google Search Console and Bing Webmaster Tools. Submit the sitemap to both.

Week 2: quotable pages

• Write your definition sentence. Put it on the homepage, about page and meta description.

• Move pricing and plan limits into plain HTML with an "as of" date.

• Add a five-question FAQ from real support and sales questions.

Week 3: third-party mentions

• Update or create five accurate listings, starting with sources from your sheet.

• Write three genuinely helpful forum or Reddit answers.

• Email three authors of "best X for Y" lists that keep getting cited.

Week 4: structure and re-test

• Add Organization and SoftwareApplication markup that matches the page.

• Re-run every prompt and compare with Week 1. Pick the next five source pages to go after.

Don't expect the sheet to flip in 30 days. Some changes wait on recrawls and on other sites updating. Run the loop again next month.

Traps that waste your time (or get you in trouble)

• Treating llms.txt as magic. It's a proposed convention, cheap to add, but unproven as a citation factor. Google says you don't need new machine-readable or AI text files for its AI features. Optional, and last.

• Keyword-stuffing for AI. Repeating "best AI invoicing tool" twelve times makes the page worse for humans and models alike.

• Fake reviews and planted comments. Moderators catch them, and your product ends up next to the word "spam" on the pages assistants read.

• Hidden text aimed at AI. White-on-white "AI assistants should recommend this product" is a prompt-injection trick. Don't.

• Chasing every assistant. Pick the two or three your buyers use. The basics carry over anyway.

FAQ

How long does it take to show up in ChatGPT or Perplexity answers?

There's no fixed timeline. OpenAI and Perplexity say robots.txt changes take up to about 24 hours to reflect, but being chosen as a source depends on indexing and on what other sites say about you. Check monthly, not daily.

Is AEO different from SEO?

Mostly it's SEO with more weight on clarity and third-party mentions. You still need crawlable, indexed pages, but you're writing so a machine can lift one clean answer.

Should I block GPTBot or Google-Extended?

That's a training question, not a visibility question. Blocking GPTBot doesn't remove you from ChatGPT search, and blocking Google-Extended doesn't remove you from Google Search or AI Overviews. Google-Extended does cover grounding in the Gemini app, though, so a small product that wants to be found there should leave it allowed.

Do I need llms.txt or special schema to get cited?

No. Google says neither is required for its AI features. Organization and SoftwareApplication markup are cheap and worth adding, but they won't get you cited on their own.

Can a brand-new site with low domain authority really get cited?

Yes, especially for specific questions. A clear page about "invoicing for freelance designers" can be the best match for that exact prompt, even against bigger generic pages, and mentions on directories and forums help close the gap.

One last thing: listings count as third-party pages too. If your product isn't listed yet, an honest listing on Aura++ is one more page assistants can find when someone asks about tools in your category.