What Do AI Assistants Say About a Brand-New Company? We Audited Our Own
By The GEO Pilot · Published
Short answer
Unless a question named us, AI assistants said nothing about our brand-new company. When GEO Pilot audited its own AI visibility in October 2026, it appeared in 0 of 45 unbranded answers, no answer linked to its site, and two assistants confused it with a similarly named plugin.
What did our AI brand audit find?
We were invisible on every question that didn't contain our name. In 45 answers to nine unbranded buyer questions, we appeared 0 times. The only mentions came from the one question that named us, and none of those answers described us fully correctly.
We launched in October 2026 and audited ourselves the same month. Here are the results:
| Metric | Result |
|---|---|
| Total answers analyzed | 50 (10 questions × 5 assistants) |
| Answers that mentioned us | 5 (10% mention rate) |
| Mentions in the 45 unbranded answers | 0 |
| Answers that linked to or used a page on usegeopilot.com | 0 of 50 |
| Mentions that described us fully correctly | 0 of 5 |
| Tone of the 5 mentions | 2 positive, 2 neutral, 1 negative |
| Share of voice | 1.1% (5 of about 454 brand mentions) |
| Average brands named per answer | 9.1 |
The headline number is 0 of 45. The 10% mention rate looks like a foothold, but it isn't one. All five mentions came from the head-to-head question where we handed the assistants our own name.
How did we run the audit?
We asked ChatGPT, Perplexity, Gemini, Claude and Grok the same 10 questions a marketing buyer might ask, once each, using our own AI visibility tracking. That produced 50 answers, and none failed.
Nine questions never named us. Examples:
- "How do I get my company's content quoted in ChatGPT and Perplexity answers?"
- "Which platform scores each piece of content for AI citation readiness and tells me exactly what to fix?"
- "Best generative engine optimization software for small B2B marketing teams on a limited budget?"
- "How do I check whether ChatGPT, Perplexity, and Gemini are mentioning my brand, and what should I change if they aren't?"
The tenth question was a head-to-head comparison naming us and a well-known AI writing tool.
Know the limits. We queried each assistant through its API with web search turned on, and results can differ from the consumer apps. Each question ran once. This is a small, single-run baseline of one brand, not an industry benchmark. Treat it as a case study, not a law.
Why did assistants confuse us with other products?
Assistants confused us because our name is shared and few sites beyond our own describe us. Claude and Grok mixed us up with an unrelated free WordPress plugin that has the same name. Grok also mentioned a third product called GEO Pilot.
Here is how the five mentions went wrong:
- Claude and Grok: described the WordPress plugin instead of our platform.
- ChatGPT and Gemini: described us mainly as an audit and measurement tool and missed that we create content.
- Perplexity: said it could not find enough independent, verifiable information about us.
This looks like an entity clarity problem, but it wasn't a gap on our own pages. Before the audit, our site, our llms.txt file and our Organization schema already shared one consistent description. Assistants still got us wrong. An assistant is more likely to trust a description it finds in several independent places, and for a new brand with a shared name, our own pages were nearly the only source. For definitions of terms such as entity, see our GEO glossary.
Why were we missing from unbranded answers?
We had no pages that answered buyer questions directly, so assistants had nothing of ours to cite. A new brand with a good product can be invisible on every unbranded question for that reason alone.
The crowding made it harder. Answers named 9.1 brands or products on average, about 454 mentions in total. Four of the five most-named "brands" were AI platforms themselves: Perplexity, ChatGPT, Gemini and Google. Together they made up about 17% of all mentions, because buyers were asking how to get cited inside those products.
The sources tell a related story. The answers surfaced 1,035 sources from 530 different domains. Average sources per answer varied widely:
| Assistant | Average sources per answer |
|---|---|
| ChatGPT | 34.8 |
| Grok | 31.5 |
| Claude | 21.8 |
| Perplexity | 11.0 |
| Gemini | 4.4 |
Reddit was the most common domain, with 81 appearances (about 8% of sources). All 81 came from ChatGPT, in 8 of its 10 answers. The other four assistants surfaced no Reddit pages. Each assistant draws on a different mix of sources, so there is no single place to show up. Our post on how AI engines choose sources goes deeper on this.
Why does this matter for a new brand?
It matters because buyers now ask assistants first, and fewer of them click through afterward. According to Google's I/O keynote in May 2026, AI Overviews has over 2.5 billion monthly users and AI Mode has passed 1 billion. TechCrunch reported that ChatGPT reached 900 million weekly active users in February 2026.
Clicks are also scarcer. Pew Research Center found that Google users clicked a traditional result in 8% of visits when an AI summary appeared, versus 15% without one. They clicked a link inside the summary in 1% of visits (March 2025 data).
If the answer is the destination, being named in it is the outcome that counts. Rankings alone no longer tell you whether that is happening, a gap we cover in rankings vs. citations. For the broader case, see why every brand needs a GEO strategy.
What is the Three Gaps model for new brands?
The Three Gaps model says a new brand is invisible in AI answers because of three separable gaps: Question, Entity and Evidence. Our audit exposed all three. Diagnose each one separately, because each has a different fix.
| Gap | What it means | What we saw | Fix |
|---|---|---|---|
| Question gap | No page answers the buyer's question directly | 0 mentions and 0 citations on unbranded questions | Publish answer-first pages that mirror real buyer questions |
| Entity gap | Your own properties don't describe you clearly or consistently | Not our gap: our site, llms.txt and Organization schema already shared one description, yet 0 of 5 mentions were fully correct | Describe the company identically on everything you control: full name, category, domain |
| Evidence gap | Few independent sources confirm who you are | Mix-ups with a plugin and a third product; Perplexity found too little verifiable information | Earn third-party mentions that describe you the same way |
Most teams jump to the Question gap because it is the easiest to act on. Our audit suggests a second lesson: a consistent description on your own properties is necessary but not sufficient. We had it, and assistants still mixed us up because few other sites describe us. For a new brand, the Evidence gap can undermine the rest.
What did we change after the audit?
We started this answer-first blog to close the Question gap: two posts a week, each answering a question buyers ask. Closing the Evidence gap, by earning mentions on other trusted sites, is the next step. We did not need to change our own description, because it was already consistent.
At GEO Pilot, the blog follows the same logic we build into the product. Each piece of content gets a 0–100 GEO score on nine signals, and we generate JSON-LD schema and llms.txt ready to publish. If you're new to the file format, read our guide to llms.txt. For scoring details, see how GEO scores work.
We haven't measured the effect yet. GEO Pilot plans to rerun the same 10 questions after this blog series and compare the results.
What should a new brand do to show up in AI answers?
Start by auditing, then close each gap. Don't assume that a consistent description on your own site is enough. Use this eight-step checklist:
- Write 10 unbranded buyer questions. Use the phrasing a real buyer would type, not your product vocabulary.
- Ask each question in several assistants. Record mentions, citations, accuracy and tone.
- Search your name. Check for products or companies that share it.
- Write one canonical description. Include your full name, what you do, who it's for and your domain.
- Use it everywhere you control. Put it on your site, in your llms.txt file, in Organization schema and on your profiles.
- Publish answer-first pages. Open each with a direct answer of 40–60 words to one buyer question. Our guide on how to get cited by ChatGPT shows the format.
- Earn third-party descriptions. Get independent sites to describe you with the same name, category and domain.
- Rerun on a schedule. Same questions, same assistants, so changes are comparable.
For the foundations, see our guides to generative engine optimization and GEO vs. SEO. If you'd rather not run the audit by hand, GEO Pilot's AI visibility tracking covers ChatGPT, Perplexity, Gemini, Claude and Grok.
Key takeaways
- We appeared in 0 of 45 unbranded answers and 5 of 50 overall. All five mentions came from the question that named us.
- 0 of 50 answers linked to or used a page on usegeopilot.com.
- None of the 5 mentions described us fully correctly. Two assistants confused us with a plugin that shares our name, and two missed that we create content.
- Our share of voice was 1.1% in answers that averaged 9.1 brands each.
- Question, Entity and Evidence gaps are separate problems with separate fixes.
- A consistent description on your own site, llms.txt and Organization schema is necessary but not sufficient. Few outside sites described us, and that is the Evidence gap.
- This is a single-run, API-based baseline of one brand, not a benchmark. GEO Pilot plans to rerun the same questions and compare.
Frequently asked questions
What is an AI brand audit?
An AI brand audit is a structured test of how AI assistants talk about your brand. You ask buyer-style questions, then record whether each answer mentions, cites and correctly describes you. Include questions that don't name you, since those show whether new buyers can find you.
How many questions do you need for an AI brand audit?
We used 10 questions across five assistants, which produced 50 answers. That is enough to expose patterns for one brand, but it is a small baseline. Rerun the same questions over time to see change.
Why would AI assistants not mention a new brand?
Assistants can only cite pages that exist and that they can verify. In our audit, we had no pages answering buyer questions directly, so we were absent from all 45 unbranded answers. A shared name and few independent sources added confusion about which company was meant.
What is entity clarity?
Entity clarity means assistants can tell exactly which company you are. It starts with describing the company the same way everywhere you control: full name, what it does and its domain. Our site, llms.txt file and Organization schema already did this before the audit, but assistants still confused us. Independent sources that describe you the same way matter too.
Do all AI assistants use the same sources?
No. In our audit, average sources per answer ranged from 4.4 for Gemini to 34.8 for ChatGPT. All 81 of Reddit's appearances came from ChatGPT, and the other four assistants surfaced none.
Make your content the source AI cites
Turn any draft, link or recording into GEO-optimized formats, scored and ready to publish. Start with $5 of free credit.
Start free