Does AI know your brand exists? Here's how to check.

Right now, someone in your market is asking ChatGPT, Perplexity, or Google's AI for a recommendation. They'll get one answer with a few names in it. Either you're one of those names, or you're not — and most brands have no idea which.

That's the part worth sitting with. BrightEdge analyzed millions of AI responses and found that 44% of prompts return zero brand mentions. Nearly half the time, the AI names nobody useful — or names your competitor and skips you. The channel that 900 million weekly ChatGPT users now treat as their starting point is the one channel almost no marketing team monitors.

You can fix that blind spot in an afternoon. This is the exact audit we run for clients, and we're handing you the whole method — the test prompts, how to read the results, where the gaps usually hide, and what to do about them. No tool required to start. Just you, the four engines, and an hour of honest looking.

Why this beats checking your Google rankings

For thirty years, "are we visible?" meant "where do we rank?" That question is now half the picture. When an AI assistant answers, there's no list to scroll. There's one synthesized answer naming a handful of sources. As we covered in how AI search decides who gets cited, getting named isn't about being the best page — it's about being the safest source to repeat. And as the gap between ranking and getting cited keeps widening, you can rank well and still be invisible in the answer.

Checking your AI visibility is no longer optional. It's the new baseline.

So here's how to do it like we do.

Four rules that separate a real audit from a vanity check

Before the prompts, the method. These four rules are the difference between data you can trust and a feel-good screenshot.

Test logged-out.

Use a temporary chat, incognito window, or signed-out session. If you query from your own logged-in account, the AI's memory of you and your personalized history will skew everything toward making you look more visible than you are. Strip the personalization. See what a stranger sees.

Run each prompt several times.

AI answers are non-deterministic — ask the same question twice and you can get different names. Run each prompt three to five times and look for the pattern, not the single result. SparkToro's research makes a sharper point worth internalizing: your position within an AI answer is essentially random noise. The metric that matters is your visibility rate — the share of relevant prompts where you appear at all.

Turn on web/search mode.

Where the platform offers it, enable live search. You want to test what the engine retrieves about you today, not a stale memory baked into its training data months ago. The two can be very different.

Test the questions your buyers actually type.

Anyone can confirm the AI knows their own brand name. The money is in the unbranded queries: "best [category] for [use case]," "who should I hire to [solve problem]." That's where buying decisions get made, and where most brands discover they're missing.

One more: write down the date. This is a moving target, and today's snapshot is a baseline you'll want to re-run.

The four tests

The audit answers four questions, in order, each one harder to pass than the last. Together they map to the three things any source needs to get cited — does the right claim about you exist, does the web agree on it, can a machine use it — which is the model we lay out in full in what generative engine optimization is.

In the prompts below, replace the bracketed parts with your own details.

The floor. Run these in each engine:

• "What is [your brand]?"
• "Tell me about [your brand]."
• "What does [your brand] do?"

What you're reading for: Does it describe you accurately? Vaguely? Does it confuse you with a similarly named company? Or does it draw a blank — "I couldn't find information about [brand]"? A blank or a mix-up here means the engine doesn't have you as a recognized entity, and nothing else in the audit will go well until that's fixed.

Knowing you exist isn't the same as describing you correctly. Probe what it believes:

• "What does [your brand] specialize in?"
• "Who does [your brand] serve, and where?"
• "What is [your brand] known for?"
• "What do customers say about [your brand]?"

What you're reading for: Correct and current — or outdated, generic, or invented? AI systems will confidently state wrong things (old locations, services you dropped, claims you never made) when the web doesn't give them a clear, consistent source of truth. Every inaccuracy here is a gap you can trace to a specific fix.

This is the one tied to revenue. Drop your brand name entirely and ask what your buyers ask:

• "What are the best [category] companies for [use case]?"
• "Who should I hire to [solve specific problem] in [location]?"
• "Top alternatives to [a known competitor]."
• "I need [outcome]. What companies should I look at?"

What you're reading for: Are you named at all? Among the first few, or only after three competitors? Never? This is the test most brands fail and never knew they were taking. If competitors get named here and you don't, that's not a branding problem — it's a citation problem, and it's costing you deals you never saw.

The source layer. Use the engines that show their sources — Perplexity, Google AI Overviews, and Gemini all surface citations; ChatGPT does in search mode. Ask the same category questions from Test 3 and read the source panel.

What you're reading for: Is your domain cited as a source? Are competitors' pages? Or is it all third parties — Reddit threads, Wikipedia, directories, review sites? If the engine is building answers about your category entirely from sources that aren't you, your content isn't in the conversation at the level that matters. That tells you exactly where the work is.

Running it across the four engines

Each engine behaves differently, so test all four and expect them to disagree — Google's own AI Overviews and AI Mode share only about 14% of their cited URLs, so a competitor dominating one tells you little about the others.

• ChatGPT. Turn on search mode; without it, you're testing months-old training memory. It leans on Wikipedia for factual questions and the live web for current ones.
• Perplexity. Use "Focus: All," and read the citations, which it shows prominently (often five or more per answer). It's the most recency-sensitive engine; fresh, recently-updated content is favored.
• Google AI Overviews. Run your buyer queries in a logged-out Google search. AI Overviews lean heavily on what already ranks organically, so this is where classic SEO strength shows up most directly.
• Gemini. Tied into Google's knowledge graph, so entity recognition (Test 1) matters even more here.

How to score what you find

Don't overthink it. For each test, in each engine, rate where you land on a simple scale: Absent → Inaccurate → Mentioned → Recommended → Cited. "Absent" is invisible; "Cited" is the win — named and used as a source. Track two numbers that actually mean something:

• Your visibility rate. Across all your unbranded buyer prompts, what percentage mention you at all? That single number is your AI visibility, and it's the one to move.
• Your relative position vs. competitors. Pick three to five direct competitors and score them on the same prompts. Your absolute rate matters less than whether you're ahead of or behind the names you compete with.

The AI Visibility Scorecard

Score your brand on each test, in each engine. Track your visibility rate, not your position.

Test / Engine

ChatGPT

Perplexity

Gemini

Google AI Overviews

01 Recognition Does it know you?

02 Accuracy Describes you correctly?

03 Recommendation Picks you, unprompted?

04 Citation Cites your content?

Worst → Best Absent → Inaccurate → Mentioned → Recommended → Cited

Your visibility rate

____% of buyer prompts mention you

Ahead of / behind

____ of your top competitors

Atomic Design — AI Visibility Audit

Where the gaps usually are — and what each one means

After running hundreds of these, the failures fall into five recognizable patterns. Each points to a different fix:

• Invisible everywhere. You barely register in any engine, even for your own name. This is an entity problem — the web hasn't established who you are clearly or consistently enough for engines to recognize you.
• Known, but described wrong or out of date. Recognition is fine, accuracy isn't. The web's information about you is stale or inconsistent, so the engine repeats the wrong story.
• Mentioned, but never recommended. You show up when asked about by name (Test 1–2) but vanish on the unbranded buyer queries (Test 3). You exist, but you're not yet the safe pick the engine reaches for.
• Competitors own the citations. Their pages are the sources (Test 4); yours aren't. Your content isn't structured or authoritative enough to be the thing repeated.
• Hallucinated facts. The engine invents details about you. This is the clearest signal of all that there's no corroborated source of truth online for it to anchor to.

What to do next

Map each gap to its fix:

• Entity / invisibility gaps → establish a clear, consistent identity across the web: a definitive description of what you are and do, published plainly and echoed everywhere engines look.
• Accuracy gaps → fix the source of truth. Correct and align the information about you across your site and the third-party sources engines trust.
• Recommendation gaps → build the authority and corroboration that make you a safe pick: genuine third-party agreement, reviews, mentions, and proof, plus content that answers the buyer's real question directly.
• Citation gaps → restructure your content to be liftable — direct answers, clear structure, real sources — so a machine can repeat it without hesitating.

That's the work of generative engine optimization, and it runs on the same Chain Reaction methodology we apply to everything: attract, impress, convert, compound.

The honest part: doing this well, continuously

The four-test audit is genuinely useful, and you can run it today for free. Run it. It'll tell you more about your AI visibility than most companies in your market know about theirs.

Here's the limit. A real, decision-grade program means a query bank of 30 to 50 buyer questions, run across four engines, sampled several times each to cut through the randomness, benchmarked against competitors, and repeated as the answers shift week to week. That's quickly thousands of checks — more than an afternoon, every month, forever.

That's the gap our AI visibility audit closes: we test hundreds of prompts across every major engine, benchmark you against your real competitors, and hand you a prioritized list of exactly which gaps to fix first. If you'd rather see your baseline without building the system yourself, that's the service. If you'd rather run it yourself, everything you need is above — go find out whether AI knows you exist.

It isn't a mystery. It's a measurement. Start with Test 1.