Your buyers are asking ChatGPT for recommendations. Here's how to find out if it names you.

A quiet shift is happening in how people find software, and most founders can't see it.

For twenty years, the job was to rank on Google. You optimized pages, earned links, and fought for the top of a list of blue links. That still matters. But a growing share of buyers now skip the list entirely. They ask ChatGPT, Perplexity, or Google's AI Overviews a direct question ("what's the best inventory tool for a small Shopify store?") and act on the handful of products the AI names back.

There is no page 2 in that answer. If your product isn't one of the names, you're invisible to that buyer, and the worst part is you'll never know. No impression, no click, no trace in your analytics. The deal simply happens somewhere you can't see.

This is what people mean by GEO, or Generative Engine Optimization: being one of the tools an answer engine actually recommends, not just a link buried on page one.

## You can measure this, and you should

Most advice about AI visibility is vibes. It doesn't have to be.

The honest test is simple: ask an AI the questions your buyers actually ask, then read the raw answer and check whether your name is in it. Not a score. Not an estimate. The literal text. You can do it by hand: open ChatGPT with web search on, ask three real category questions, and see if you're named and who is named instead.

I built a free tool that automates exactly that, because I wanted the answer for my own products and didn't want to run it manually every time. You paste your domain, it sends three genuine buyer questions to a live model, and it shows you the raw result: are you mentioned, in how many of the three, and which competitors got recommended. No account, no charge.

I'll be honest about a mistake I nearly shipped. The first version leaned on questions that already contained the brand name, so almost everyone "passed." Good for conversions, useless as a signal, and dishonest. I tore it out and made the questions brand-free and the detection binary. When I tested it against a made-up brand, it correctly returned "not visible" instead of inventing a flattering number. That's the bar I want.

## Why most products are invisible in AI answers

If you rank on Google but never show up in ChatGPT, you're not failing at SEO. GEO is a different surface with a different mechanic.

Answer engines lean on structured, consistent, citeable sources. When a model synthesizes a recommendation, it pulls from what it can quote confidently: facts that appear the same way in multiple places, are easy for a machine to parse, and come from sources beyond your own marketing homepage.

The products that get named tend to share three traits:

- Consistent naming everywhere. The same product name on the site, in directories, in reviews, in roundups. Mixed naming confuses attribution.

- Third-party listings and mentions. Directories and review sites act as corroborating sources a model can cite. A fact stated only on your own site is weaker than the same fact echoed independently.

- Machine-readable facts. Clear pricing and plain feature descriptions, not buried in an image or a video.

## What to actually do

You can't edit a model's training data, but you can shape what the next crawl sees and what the model can cite today.

1. Fix your naming. One canonical name, used identically everywhere. Free, and the most common own-goal.

2. Get onto structured third-party listings so answer engines have corroborating sources.

3. Publish machine-readable facts: plain pricing and features, structured data where you can.

4. Exist in the comparison and "best X" content models pull from. You don't need to game it, you need to be in it.

5. Re-check. GEO isn't set-and-forget. Baseline now, fix, re-check in a few weeks.

One concrete step that touches most of that at once is a structured, honest directory listing. That's what I'm building with directree, the honest software directory: paste a URL, get a full structured listing in about 30 seconds, every field labelled Observed, AI-inferred, or Founder-edited, with no fake precision scores. Because the data is structured and honest, it's the kind of source answer engines lean on, and claiming your listing gets you a do-follow backlink too. It won't single-handedly make ChatGPT recommend you (nothing will), but it adds a machine-readable, corroborating source about your product to the open web, which is exactly what GEO rewards.

If you've built something, take a minute and check whether the AI actually names you. It's free, and knowing beats guessing.

Check your GEO visibility: https://www.directree.io/geo-check

Read the full playbook: https://www.directree.io/blog/is-chatgpt-recommending-your-saas