Try something right now.

Open ChatGPT, Perplexity, or Gemini and type in a query your ideal customer would ask before buying your product.

Something like: "What's the best project management tool for a 15-person startup?"

Look at the answer. Not the ads. Not the search results. The answer.

If you're not in it, your potential customers aren't finding you in the fastest-growing discovery channel in the world. And unlike traditional search, there's no page 2 to scroll to. There's no page at all.

AI gives them three to five options. They pick from the list. If you're not on it, you don't exist.

How AI search actually works (and why it's nothing like Google)

Traditional SEO trained us to think about rankings. Position 1 through 10. Blue links. Click-through rates.

AI search doesn't rank. It recommends.

When someone asks ChatGPT for a project management tool, it doesn't crawl the web in real time and score pages. It synthesizes everything it's absorbed about your brand, your competitors, your reviews, your content, your presence across the internet, and generates a curated shortlist.

This means the game has changed in three fundamental ways:

1. It's a reputation game, not a ranking game. Google rewarded pages. AI rewards brands. The model's "opinion" of you is built from thousands of signals across your entire web presence, not a single optimized page.

2. Every model sees a different internet. ChatGPT, Perplexity, Gemini, and Claude each train on different data, weight sources differently, and update on different cycles. A study of 100,000 prompts across ChatGPT and Perplexity found only 11% overlap in the domains they cited. You might be visible in one model and invisible in another.

3. Size matters less than you think. In traditional search, domain authority gave big brands an almost unbeatable advantage. In AI search, the playing field is more open. Models care about consistency and specificity. A startup that shows up clearly and consistently across niche sources can outperform a Fortune 500 company that has a scattered, generic presence.

This is the window. It won't stay this open forever.

The 5-step AI visibility check

Here's a practical way to find out where you stand. This takes about 20 minutes.

Step 1: Ask what your customers ask

Open ChatGPT, Perplexity, and Gemini. In each one, type 5–10 queries your ideal customer would ask before buying. Not branded queries (those are easy). The unbranded ones:

• "Best [your category] for [your audience]"
• "How do I solve [problem your product solves]?"
• "[Your competitor] vs alternatives"
• "What tools do [your target customer] use for [job to be done]?"

Track your results in a simple grid: query on one axis, model on the other. Mark whether you appear, and in what position.

Step 2: Test across all major models

Don't stop at one. The 11% domain overlap finding means your visibility is model-specific. You might rank #1 in Perplexity and not exist in ChatGPT.

Test at minimum:

• ChatGPT
• Perplexity
• Gemini
• Claude

You're looking for patterns. Where are you consistently present? Where are you consistently absent? Where does it vary?

Step 3: Study the descriptions (this is where the real insight is)

This is the step most people skip, and it's the most valuable.

Don't just check whether you appear. Read how the model describes you.

When AI recommends your product, it attaches a description. That description is your brand positioning as understood by the model. You didn't write it. You don't control it directly. But it determines whether the user clicks or scrolls past.

Ask yourself:

• Does the description match your actual positioning? If you're a premium product and the model describes you as "a budget-friendly option," you have a perception problem.
• Is it specific or generic? "A popular project management tool" tells the user nothing. "Built for async-first remote teams with deep Notion integration" gives them a reason to choose you.
• Does it mention your key differentiator? If the model doesn't know what makes you different, neither will the user.

Compare your descriptions to your competitors' descriptions. If theirs are more specific and compelling, the model has a better understanding of their positioning than yours. That's fixable, but you have to know it's happening first.

Where descriptions come from: The model builds its understanding from your website copy, review sites, Reddit threads, press coverage, comparison articles, and documentation. If those sources tell an inconsistent story, the model's description will be vague or wrong. If they tell a clear, consistent story, the description will be sharp.

Step 4: Identify why you're winning or losing each query

For every query where a competitor appears and you don't (or where their description is stronger), diagnose the gap.

It's usually one of three things:

Coverage gap: The competitor has more content addressing that specific query across more sources. They have a blog post, a comparison page, Reddit comments, a G2 review, and a podcast mention all reinforcing the same message. You have a homepage and a blog post.

Consistency gap: You have the content, but it tells different stories in different places. Your homepage says "enterprise-grade." Your Product Hunt launch said "perfect for solo founders." Your G2 reviews focus on price. The model can't form a clear picture because your signals conflict.

Specificity gap: The competitor is positioned for a specific niche. You're positioned for everyone. AI models favor specificity because it makes the recommendation more useful. "The best CRM for B2B SaaS companies with 10-50 employees" beats "a CRM for businesses of all sizes."

Once you know which gap is the problem, you know where to focus.

Step 5: Prioritize by buying intent

You now have a map of where you're visible, how you're described, and where the gaps are. The question is: where do you start?

Start with the queries closest to a purchase decision. Here's how to identify them:

High buying intent (fix these first):

• "[Category] vs [category]" comparisons
• "Best [tool] for [specific use case]"
• "[Competitor] alternatives"
• "Which [category] should I use for [specific job]"

Mid intent (fix these second):

• "How to [solve problem your product addresses]"
• "What tools do [role] use for [job]"

Low intent (fix these later):

• Broad educational queries
• Industry trend questions

For your top 2-3 high-intent queries where you're missing or poorly described, build a focused plan: what content needs to be created, updated, or distributed to close the coverage, consistency, or specificity gap you identified in Step 4.

Why we built Saturation

We built Demand Curve to help startups operate at the frontier of growth. That hasn't changed.

But when we look at AI, we don't see a new channel to bolt onto the existing playbooks. We see a platform shift that requires its own approach. New mental models, new skill sets, new team structures.

So we took our own advice and built something purpose-built for it. That's Saturation.

If you want to see how your startup shows up in AI search today, start with a free AI search audit at saturation.co →​

This is Part 1 of a three-part series on AI search strategy.
‍Next up: How to systematically build your AI visibility across models.