Why your brand is invisible in ChatGPT (and the diagnostic to fix it)

Most founders have never run the basic diagnostic on their own product.

They check Google rankings. They look at their Search Console traffic. They watch their backlinks. But they have never opened ChatGPT and asked it "what is the best tool for X" to see whether their product shows up. They do not know whether Perplexity has even heard of them. They have no idea what Claude says when someone asks about their category.

This is a problem because buyers now research with AI assistants before they touch Google. If your product never surfaces in those answers, you do not exist for that customer. The buyer makes a shortlist, sends three vendor inquiries, and you are not on it. You never get the chance to compete.

This post walks through the diagnostic question that matters most in 2026: do AI engines know what your product does, and do they say it accurately? You can run this audit manually in fifteen minutes. We will also show you how the AfterLaunch snapshot runs it in sixty seconds across four engines.

How AI engines actually source content

Before you audit, it helps to understand the mechanism. AI engines do not have a single way of finding information about you. They have several, and the mix differs by engine.

The crawler-citation pipeline

GPTBot, ClaudeBot, PerplexityBot, and Google-Extended each crawl the open web continuously. Their crawl footprints overlap heavily with traditional search crawlers but with different access patterns. ChatGPT cites sources in 87% of responses according to recent measurement. Google AI Mode cites 76% of the time. Google AI Overviews cite 85% of responses. Perplexity visits roughly 10 pages per query and cites three or four of them in its answer.

There are two distinct pipelines feeding citations. The first is training-data influence: information absorbed during model training that the model can recall without web access. The second is live-retrieval citation: information fetched during the query itself and cited inline. Most public-facing AI products mix both. Perplexity leans almost entirely on live retrieval. ChatGPT leans heavily on training data with web search activated on demand. Claude sits somewhere between.

The practical implication: if your brand was not in the training corpus at scrape time, you are invisible to that engine until either the next training cycle, or until the engine fetches a live page about you. The fastest path to visibility is producing live, citable content that the retrieval layer can grab on demand.

What "cited" actually means

Citation, mention, and recommendation are three different things. Mixing them up will confuse your audit.

• Mention. The AI answer references your product by name but does not link to your domain. The buyer sees the name but cannot click through to verify or learn more.
• Citation. The AI answer includes a source link, often footnoted, that points to a page on your domain or a page that talks about you. The buyer can click through.
• Recommendation. The AI answer affirmatively suggests your product as a solution to the buyer’s question. The strongest possible outcome.

Some answers cite without recommending (your page is a footnote in a list of options). Others recommend without citing (your name appears but no link). The signal you want is positive recommendation with citation. This is the rarest and the most valuable.

Why some brands win citations and others do not

Five factors keep showing up in the data. Domain authority is the largest single predictor: domains with rating above 50 appear in AI answers roughly five times more often than domains below 30. But domain authority alone does not explain everything. Plenty of high-DR domains get ignored. Plenty of low-DR domains punch above their weight.

The other four factors that consistently matter:

• Cross-platform authority. Brands cited across Reddit, G2, Capterra, Hacker News, and mainstream press get more AI citations than brands cited only on their own domain. AI engines look for agreement across independent sources before citing confidently.
• Structured data presence. Pages with FAQPage schema, HowTo schema, and Organization schema get extracted more cleanly into AI answers. AI models often quote tables and structured lists almost verbatim.
• Content depth and specificity. Pages with concrete numbers, bounded claims, and self-contained sections get absorbed into AI answers more easily than vague, hedged prose. "Most companies see improvement" is unhelpful. "Citation rates of 20% to 30% indicate healthy visibility" is citable.
• Recency. Content updated regularly gets cited more often than content that has sat untouched for two years. Perplexity in particular weights recency heavily. Most engines penalise stale content.

Notice what is missing from this list. Pretty design. Long-form content for its own sake. Keyword density. Backlinks from low-quality sites. The signals AI engines weight are different from the signals classic SEO tools score.

The audit framework

Three checks. Roughly fifteen minutes if you do them manually. Cover the categories of failure that matter.

Surface check: direct prompts

Open ChatGPT. Type these three prompts, swapping in your product name:

• "What is [your product]?"
• "Tell me about [your product]."
• "Is [your product] a good tool for [your category]?"

Repeat in Perplexity. Repeat in Claude. Repeat in Google AI Mode.

You are looking for three outcomes. Best case: each engine describes your product accurately and links to your site. Mid case: each engine knows of you but describes you wrongly or with outdated information. Worst case: one or more engines have never heard of you, or worse, confuse you with another product.

The worst case is more common than founders expect. From the patterns we are seeing as founders run the snapshot, roughly one in three SaaS products has a confusion problem on at least one engine. ChatGPT cheerfully describes a product that does not exist under your brand name. Perplexity links to a competitor’s page when asked about you. These errors stay in place for months unless you actively work to correct them.

Authority check: category prompts

The surface check tells you whether AI engines know you exist. The authority check tells you whether they consider you when buyers ask category questions, which is where most discovery actually happens.

Pick five questions your buyer would actually type:

• "Best tools for [problem you solve]"
• "How do I [outcome your product delivers]"
• "Alternatives to [your largest competitor]"
• "[Category] tools for [your buyer’s context, e.g. solo founders, B2B sales, indie hackers]"
• "Recommend a tool for [specific job to be done]"

Run each in all four engines. Record whether you appear in the answer set. If you appear, record your position in the list. If you do not appear, record who does.

The benchmark to know: for B2B SaaS in 2026, citation rate below 10% across your category prompts means you are functionally invisible. The healthy band is 20% to 30%. Above 40% is category leadership. Seed-stage startups realistically start at 2% to 8% and climb from there over months.

If your authority check is empty across the board, that is your biggest growth problem. It is also the most fixable one. Authority builds through compound mentions, not through more blog posts on your own domain.

Voice check: sentiment of mentions

The third check is the one founders skip and later regret. When AI engines do mention you, what do they actually say?

Read the descriptions carefully. Are they accurate? Are the claims out of date? Are they neutral, positive, or negative in tone? Are they comparing you favourably or unfavourably to competitors? Are they describing the product you sell today, or the product you sold eighteen months ago?

Sentiment problems are easier to fix than surface problems. A wrong description usually traces back to one or two pages: an outdated review on a comparison site, a misleading G2 entry, a year-old Reddit thread that ranks for your brand name. Identifying these and either updating the source or producing fresher counter-content typically shifts the sentiment within a few weeks.

Surface problems are harder. If ChatGPT has no idea you exist, you need to do the underlying work of being mentioned across the web until the next training cycle picks you up. That is months of consistent presence, not days.

Running the diagnostic

The manual version above is the slow version. It takes fifteen minutes and tells you the basics. It is enough to identify whether you have a surface problem, an authority problem, or a voice problem. For most founders, that is the most important thing to learn.

The faster version is what AfterLaunch was built to do. The free snapshot runs the same three checks automatically against ChatGPT, Perplexity, Claude, and Google AI Overviews, plus a few additional signals. It returns the diagnostic in sixty seconds with no signup.

Here is what to look for when you read the snapshot output:

• AI Visibility section. Citation chips per engine. If you see "Not mentioned" on more than two engines, you have a surface problem to fix first.
• Sentiment chips. Positive, neutral, or negative for each citation. Lots of neutral citations means engines mention you but do not recommend you. That is recoverable.
• Competitive landscape. Which competitors are cited when you are not. This is your benchmark group for compound-mention work.
• Five actions. The snapshot ends with five concrete actions ordered by leverage. Start with the highest one. The rest can wait.

The point of the diagnostic is not the score. The point is to know which problem you are solving. Founders who skip this step usually spend months on the wrong work. They publish more blog posts when their actual issue is that AI engines have never seen their brand on any third-party site. They build backlinks when their actual issue is that the brand voice in existing mentions is wrong.

Diagnostic first. Then action. Otherwise you are just guessing.

Where to go from here

If you have not done the manual audit, do it before you do anything else this week. Fifteen minutes. Open ChatGPT, run the three direct prompts, repeat across Perplexity and Claude. The picture you get will tell you what to spend the next quarter on.

If you want the automated version that runs the four-engine comparison and turns it into concrete actions, that is what AfterLaunch does. Join the waitlist for early access.

Once you have your diagnostic, two reads pair well with it.

The strategic context: why organic growth changed in 2026 and what the new playbook looks like.

The execution detail: the specific best practices and pitfalls that determine whether AI engines actually cite you, based on patterns we have seen across hundreds of scans.