How I Got My SaaS Mentioned by ChatGPT: A Founder's AEO Playbook

Somewhere around six months ago, one of my SaaS products started getting signups I couldn't trace to any channel. No ad, no search query, no referral link. I eventually figured out what was happening: people were asking ChatGPT or Perplexity for a recommendation in my category, getting my product's name back, and just going straight to the pricing page. No funnel, no click trail, just a cold conversion.

That's a good problem to have, but it also means the old playbook for getting found doesn't fully apply anymore. Here's what I actually did to go from "no idea if AI assistants know I exist" to getting mentioned with some regularity, and how I keep tabs on it now.

Step 1: find out where you actually stand

Before doing anything else, go ask the models yourself. Open ChatGPT, Claude, Perplexity and Gemini and type the exact questions a buyer would type: "best AI translator for Mac," "tools to track earthquakes in real time," whatever fits your category. Not your brand name, the problem your buyer has.

Do this five or six times per model, not once. Answers aren't deterministic. You'll get different orderings, sometimes a different set of tools entirely. One run tells you nothing. A handful of runs starts to show you a pattern: are you showing up at all, and if so, where in the list and next to which sources.

This manual pass is free and takes twenty minutes. It's also exactly what a tool like AskAiRank automates once you're doing it often enough that copy-pasting prompts into four chat windows gets old.

Step 2: figure out what the model is actually citing

This is the part that matters most and gets skipped the most. When a model recommends a tool, it's usually because it read something, a comparison page, a review site, a Reddit thread, a "best of" listicle, your own docs. Look at the sources cited in the answer (Perplexity and the citation-heavy modes make this easiest to see).

In my case, I noticed one of my competitors kept getting cited from a specific third-party comparison article, over and over, across different prompts. That article had a clean table: features, pricing, pros and cons, one line per competitor. My own site had none of that. Guess which page the model preferred quoting from.

Step 3: build the kind of content models actually lift from

Once you know what's getting cited, the fix is usually boring and mechanical:

• A real comparison page. Not marketing fluff, an honest table against two or three named competitors, with pricing and feature differences spelled out plainly. Models seem to love pulling structured comparisons because they're easy to summarize.
• A clear "what this does" page near your homepage, written in plain language, not jargon. If a model has to guess what your product does from vague copy, it'll skip you for a competitor who spelled it out.
• Pricing that's actually visible. If your pricing lives behind a "contact us" wall, models can't cite a number, and buyer-intent questions almost always include a price angle.
• Documented facts, not adjectives. "Free plan, 10 prompts, no card required" gets quoted. "Industry-leading" doesn't.

Step 4: get into the places models already trust

Models weight some sources more than others, and you can't fake your way onto a "best of" listicle you don't belong on. But you can earn your way in:

• Submit to relevant directories and listicles in your category (the honest ones, not link farms).
• Show up in Reddit threads where people are actually asking your category's version of "what should I use for X." Answer like a person, not a billboard.
• Get reviewed somewhere with real detail, not just a badge. A three-paragraph review beats a logo wall every time for citation purposes.

This part takes longer than writing a comparison page, and there's no shortcut. It's closer to old-school link building than to prompt engineering.

Step 5: measure whether any of it worked

Here's where the manual approach from Step 1 gets tedious fast. If you're only checking occasionally, you can't tell noise from a real trend. I started running the same prompt set on a schedule and tracking a visibility score over time instead of eyeballing individual answers, which is the whole reason I ended up building AskAiRank rather than just living with a spreadsheet of screenshots.

Whatever you use to track it, the thing to watch isn't a single "yes I got mentioned" moment. It's the trend over three or four weeks after you ship a comparison page or land a new citation-worthy mention. If your visibility inches up after that, you know the content actually mattered. If it doesn't move, you learn that too, and you go back to Step 2.

The honest summary

Nobody has this fully figured out yet, including the tools charging $300/month for it. The mechanics that seem to actually move the needle are unglamorous: know what gets cited, build the page that deserves to be cited, get mentioned in places models already trust, and check back regularly instead of once. It's slower than a growth hack and it actually works.