Across 2,545 AI answers collected over eleven weeks, ChatGPT, Gemini, Perplexity and Claude named our parent company, Nurdd, in 4 of the 2,166 answers to category questions that did not mention it by name, while naming a tracked competitor in 668 of them.
We build Noma by Nurdd, which asks AI engines the questions a brand’s buyers ask and records what they answer. We track our own company with it, the same way a customer would. This post publishes those counts. It is our data about our brand, not a benchmark: every figure comes with the sample it was counted on, and the limits are listed at the end.
What we measured
| Item | Value |
|---|---|
| Brand | Nurdd (marketing technology) |
| Engines | ChatGPT, Gemini, Perplexity, Claude |
| Dates | 6 August to 22 October 2026 |
| Prompts | 96: 85 that do not name Nurdd, 11 that do |
| Answers | 2,545 completed answers |
A prompt that names the brand (“What is Nurdd and what does it do”) tests whether the engines know you. A prompt that does not (“Best Influencer Marketing tools in India”) tests whether they recommend you, and that is the one buyers ask. We explain how to choose them in which prompts to track.
Finding 1: the engines know Nurdd but almost never recommend it
When the prompt included the name, the engines named Nurdd in 377 of 379 answers (99.5%). When it did not, they named Nurdd in 4 of 2,166 answers (0.2%). Being known and being recommended are different results, and only the second shows up when a buyer asks about the category.
Finding 2: competitors were named where we were not
At least one of the competitors we track was named in 668 of the same 2,166 category answers (30.8%). The engines were answering these questions with brands in them; ours was rarely one. That gap is what AI share of voice measures.
Finding 3: two engines nearly always list sources, two often do not
| Engine | Answers | Listed sources |
|---|---|---|
| Perplexity | 617 | 617 (100%) |
| ChatGPT | 657 | 636 (96.8%) |
| Gemini | 646 | 324 (50.2%) |
| Claude | 625 | 246 (39.4%) |
Where an engine lists sources, you can see which pages shaped the answer and work on them. Where it does not, the answer is all you have. We cover why the engines differ in why AI engines disagree.
Finding 4: answers draw on many sources, and rarely ours
The 1,823 answers that listed sources cited a median of 7 each, from 2,247 different domains in total. Nurdd’s own site was among them in 106 answers (5.8%). Even for a question about your own category, your site is one source among many, and usually not one of them.
Finding 5: a mention rarely comes with a citation
Of the 381 answers that named Nurdd, 30 (7.9%) also cited its website. A mention and a citation are separate results, as we explain in citation vs brand mention; for us the first came far more often than the second.
Finding 6: the tone was mostly neutral
Of the 381 answers that named Nurdd, 252 (66.1%) described it neutrally, 87 (22.8%) favourably and 42 (11.0%) unfavourably. A neutral mention says what a company does; a favourable one recommends it. Noma scores each mention with the passage it came from, described in brand sentiment in AI answers.
Finding 7: marketing questions are answered from marketing tool sites
The domains cited most often, as a share of the 1,823 answers with sources, were semrush.com (17.4%), analytics.google.com (15.9%), ahrefs.com (12.7%), support.google.com (11.6%) and linkedin.com (9.4%). In this category, the engines leaned on large vendors’ guides and documentation. Those are the pages a mention would have to come from, which is the case for earning coverage elsewhere rather than only publishing on your own site.
What we are doing about it
- Answering category questions directly. Pages that lead with a one-sentence answer to the question a buyer asks, as described in how AI engines choose citations.
- Earning mentions on the sites the engines cite. Listings, reviews and comparison pages in the category, because the answers are built from them.
- Measuring the same 96 prompts on a fixed schedule. So a change in these counts means something changed, not that the questions did.
Limits of this data
- One company, in one category. Other brands will see different numbers.
- 96 prompts that we wrote. A different set would give different counts.
- Four engines, asked the way Noma asks them; an answer in someone’s own app can differ.
- Eleven weeks, from 6 August to 22 October 2026. Engines change their models and search often, so these are a snapshot.
- No customer data is included; the counts are from tracking our own brand only.
To see the same counts for your brand, run a free brand audit.
See what AI engines say about your brand
Free · no account · or start a 3-day trial
Questions people ask
How often did AI engines name Nurdd when asked about its category?
In 4 of 2,166 answers (0.2%) to prompts that did not include the brand name, across ChatGPT, Gemini, Perplexity and Claude between 6 August and 22 October 2026. When the prompt included the name, the engines named Nurdd in 377 of 379 answers.
Which AI engines list their sources most often?
In this sample Perplexity listed sources in all 617 of its answers and ChatGPT in 636 of 657 (96.8%). Gemini listed sources in 324 of 646 answers (50.2%) and Claude in 246 of 625 (39.4%).
How often was Nurdd's own website cited?
Nurdd's site was among the sources in 106 of the 1,823 answers that listed any (5.8%). Of the 381 answers that named Nurdd, 30 (7.9%) also cited its site.
Is this an industry benchmark?
No. It is one company's tracking, 96 prompts in marketing technology, on four engines over eleven weeks. It shows what the engines did for this brand, with every count stated, not what they do for every brand.