A founder we worked with had a homepage that said “chatbot” and “voice bot” in the header. Somewhere along the way, a rebrand exercise decided those words sounded dated, so they got swapped for “intelligent automation platform.” Cleaner. More ambitious-sounding. Within weeks, the company had dropped out of Google rankings for the exact terms their buyers searched, and out of ChatGPT’s answers to the same questions. Nobody had touched the product. They’d just made themselves unrecognizable to the systems buyers actually use to find companies like theirs.

That’s the trap under most “AI visibility” advice right now. Founders assume the problem is exotic, some new technical layer they need a tool to decode, when the more common problem is that they’ve made themselves harder to describe, in language, in structure, or in positioning, and no AI system can recommend what it can’t parse.

Before you pay for a monitoring tool, run the checklist below yourself. It takes about fifteen minutes and tells you more than most audits do, because you already know your buyers better than any tool does.

Why most AI visibility advice skips a step

A lot of the content on this topic treats AI search as disconnected from traditional SEO, as if ChatGPT and Perplexity invented their own independent ranking systems. They didn’t, mostly. Traditional SEO still drives the large majority of what AI answer engines cite, because these tools still lean heavily on Google’s index to generate their answers. Schema markup and structured data help retrieval, but they don’t substitute for a site that’s actually clear about what you do.

That matters for the checklist below, because it changes the order of operations. If you’re invisible in AI search, the fix usually isn’t a new “AI content” strategy bolted onto your existing site. It’s making sure the underlying pages, the category language, the use cases, the comparisons, are explicit enough for a model to quote. Run the audit first. Most founders find the gap is upstream of AI entirely.

The checklist

Run this in a logged-out or private browser session wherever possible. AI tools personalize based on your history, and your own search behavior will quietly skew what you see.

1. Write down the prompts your buyers actually type, not your brand name

Testing your own company name tells you almost nothing useful. Write ten to fifteen prompts across three categories instead:

  • Category prompts. “Best [your category] for [specific buyer type],” “who should I use for [the job you do].”
  • Comparison prompts. “[You] vs [named competitor],” “is [competitor] worth it,” “alternatives to [category leader].”
  • Problem-first prompts. The question a buyer types before they know your category exists, worded the way they’d actually type it, not the way your positioning deck words it.

If you’re not sure what your ICP actually types, this is also a useful moment to check whether you’ve done that homework at all. It’s a more common gap than founders expect.

2. Run every prompt across all four major engines separately

Treat ChatGPT, Perplexity, Google AI Overviews (or AI Mode), and Copilot as four different channels, not one. It’s common to see strong presence in one and near-total absence in another, the same company can read as a well-cited authority to ChatGPT and be functionally invisible to Perplexity, because the two draw from different retrieval and ranking logic. If you only test one engine, you’ll draw the wrong conclusion about your overall visibility.

3. Score each result honestly

For every prompt, in every engine, log one of four outcomes:

  • Cited. Your site appears as a linked source backing the answer. Best case.
  • Named. You’re mentioned by name in the answer, with or without a link.
  • Paraphrased. Your ideas or positioning show up unattributed, the model learned from you but doesn’t credit you.
  • Absent. You don’t appear at all, even on prompts where you’d reasonably expect to.

A spreadsheet with one row per prompt and one column per engine turns this from a vague impression into a real baseline you can track over time.

4. Check whether you’ve accidentally erased your own category language

This is the mistake from the opening example, and it’s more common than it should be. Founders trying to sound more sophisticated often strip out the plain, searchable words for what they actually do, “chatbot” becomes “conversational intelligence layer,” “email tool” becomes “lifecycle orchestration platform,” in favor of language that sounds impressive to a human skimming a homepage but tells an AI system nothing concrete to match against a buyer’s question. If your product pages don’t state your category in the same plain words your buyers use, you’re asking the model to guess. It usually guesses toward whoever did say it plainly.

5. Check whether you show up in the comparison and listicle content AI tools cite

A lot of AI answer-engine output for “best X for Y” questions draws directly from existing comparison articles and roundup listicles, the AI isn’t independently evaluating every vendor, it’s often summarizing what a listicle already said. If your company is absent from the third-party roundups and comparison posts your competitors appear in, that absence carries straight through to AI answers. Search for the comparison and “best [category]” content that exists in your space and check whether you’re in it. If you’re not, that’s a distribution gap worth closing directly, through outreach to the people who maintain those lists, not just through your own site.

6. Confirm the fundamentals underneath are actually solid

Once you know where the gaps are, check whether the underlying page-level basics are in place: a clear, explicit statement of your category and who you’re for, use-case pages in buyer language, honest comparison pages against named alternatives, and specific, verifiable proof (named customers, real numbers) rather than adjectives. These fundamentals are what make a site legible to AI in the first place, and what makes your claims something a model can actually quote back accurately rather than paraphrase into something vague or wrong.

7. Re-run the whole thing in 30 days

A single pass gives you a snapshot, not a trend. Models update, indexes refresh, and competitors change their own content, so what’s absent this month may be present next month, and vice versa. Set a recurring calendar reminder, not a one-off task, so you actually have before-and-after data instead of a single data point you can’t act on.

A worked example

Take a hypothetical B2B SaaS company selling inventory forecasting software to mid-market e-commerce brands. Running the checklist, they find they’re cited reasonably well in ChatGPT for “best inventory forecasting software for Shopify brands,” but absent entirely from Perplexity for the same prompt, and absent from both for “how do I reduce stockouts without over-ordering,” a problem-first prompt closer to how their actual buyers start their search.

Digging into why: their homepage leads with “AI-powered supply chain intelligence platform,” a phrase that appears nowhere in how buyers describe the problem. Their use-case pages don’t mention “stockouts” or “over-ordering” anywhere in the text, even though that’s the exact language in their own sales calls. And they’re absent from two prominent “best inventory management tools” roundups that competitors are cited in.

None of that required a paid monitoring tool to find. It required running the checklist and being honest about the gap between the language on the site and the language buyers actually use.

Common mistakes

Assuming Google rankings carry over automatically. They influence AI visibility heavily, but engine-specific retrieval and citation behavior means a page ranking well in Google can still be skipped over in an AI-generated answer, particularly on Perplexity, which behaves differently from Google-index-reliant tools.

Testing only your brand name. This confirms you exist. It tells you nothing about whether you show up when a buyer who’s never heard of you asks the category question.

Treating every engine as one channel. The most common blind spot in this checklist. Strong ChatGPT presence and weak Perplexity presence (or the reverse) is normal, not an anomaly, and averaging them together hides exactly the gap you’re trying to find.

Chasing sophisticated-sounding language over plain category words. Vague, ambitious-sounding positioning reads well in a pitch deck and reads as nothing to a model trying to match your page to a buyer’s question. Say what you are in the words your buyers already use, then layer the ambition on top.

Buying an audit tool before doing the manual pass. Paid AI-visibility monitoring tools are genuinely useful once you’re tracking dozens of prompts on an ongoing basis. They’re not necessary to get your first real baseline, and starting with one means you skip the step of writing your own buyer-intent prompts by hand, which is itself a useful exercise in understanding how your buyers actually think.

The takeaway

AI visibility isn’t a separate discipline that requires an entirely new playbook. It’s mostly a mirror held up to positioning clarity you either have or don’t. The founders who show up in AI answers are, almost without exception, the ones who state plainly what they do, who it’s for, and why they win, in the same language their buyers already use. Run the checklist, find out honestly where you stand, and fix the positioning gap before you fix anything technical.

If you want a second pair of eyes on where your GTM positioning is creating this kind of blind spot, get in touch.