We Ran an AI Visibility Audit on Nike, Adidas, Puma and Under Armour. Here's What AI Got Wrong
On 15 October 2026 we ran an AI visibility audit on four of the biggest names in sportswear: Nike, Adidas, Puma and Under Armour. We asked AI platforms the questions shoppers actually ask, recorded who they named, who they skipped and who they linked to, then checked the facts in them by hand against primary sources.
Two things came out of it. AI rarely puts Under Armour at the top when a shopper asks for the best gear. And when it does talk about Under Armour, one platform says something that has not been true since 2020.
How we ran the audit
Five questions a shopper asks without naming a brand:
- best running shoes 2026
- best basketball shoes
- best soccer cleats
- best workout clothes for the gym
- most popular sportswear companies
Then three questions per brand that do name it: who is the CEO, what is its history, and how it compares with a rival. That made 67 answers in total.
| Platform | Model | How it answered |
|---|---|---|
| ChatGPT | gpt-5.6-luna | From memory, web access off |
| Perplexity | sonar | Live web search |
| Claude | claude-sonnet-5 | From memory, web access off |
| Gemini | gemini-3.6-flash | Grounded search. Left out of the visibility counts below: a token limit on our side cut its answers short in this run, and a cut-off answer is not a fair "no" |
Two limits, stated up front. Five questions are a sample, not the market: AI platforms rewrite a question before they search, so no fixed set of prompts sees exactly what a shopper sees. And this run did not include Google AI Overviews. We are not affiliated with any of the four brands.
Who AI names first: share of voice across 15 answers
These counts cover the five unbranded questions on ChatGPT, Perplexity and Claude. That is 15 answers, and every brand is measured on the same 15.
| Brand | Mentioned | Named first | Share of brand mentions | Linked its own site |
|---|---|---|---|---|
| Nike | 14 of 15 | 13 | 45% | 1 |
| Adidas | 13 of 15 | 2 | 34% | 0 |
| Puma | 9 of 15 | 0 | 12% | 0 |
| Under Armour | 9 of 15 | 0 | 8% | 0 |
Nike was named first in 13 of 15 answers. Under Armour was mentioned in 9 and named first in none. An AI answer reads like a ranked list, and Under Armour never led one. Being mentioned and being recommended are different results, so an AI visibility audit has to record both.
The question with the most demand
"Best running shoes 2026" is searched about 4,400 times a month in the US (Google Ads search volume via DataForSEO, pulled 16 October 2026). Of the four brands, ChatGPT named Nike, Adidas and Puma, in that order. Perplexity and Claude did not name Under Armour either. On the highest-demand question in the set, Under Armour was absent from all three platforms.
Two real AI hallucination examples
We checked the answers by hand against primary sources, starting with the facts that matter most to a brand: ownership, leadership, founding and headquarters. Most were right: all 16 answers to "who is the CEO" named the correct current CEO. We found three wrong statements, and they share a pattern.
1. "UA owns popular fitness-tracking apps"
Asked "Under Armour vs Nike", Claude listed Under Armour's advantages and wrote:
"Connected fitness (MyFitnessPal, MapMyRun) – UA owns popular fitness-tracking apps, giving it a data/engagement edge Nike doesn't fully match"
Under Armour owns neither app. It completed the sale of MyFitnessPal to Francisco Partners on 18 December 2020 (SEC filing), and Outside Inc. acquired MapMyFitness, including MapMyRun, from Under Armour on 30 August 2024.
The strange part: Claude knew. In a different answer about the same brand, the same model wrote that connected fitness was "largely divested — MyFitnessPal was sold in 2020". The error was not a gap in what the model knew. It changed with the question.
2. Two different years for one CEO change
Asked "who is the CEO of Puma", Claude named Arthur Hoeld, correctly, and said he took over in 2024. Asked for Puma's history, it said 2023. PUMA's own management page lists him on the Management Board as CEO since July 2025. Right name, wrong year, and a different wrong year depending on the question.
For more cases outside this audit, including two that ended in a tribunal and a courtroom, see our list of AI hallucination examples.
What the errors have in common
AI gets wrong what changed. A sale and a leadership change. The stable facts came back right. The errors sat in facts with a before and an after.
One likely reason: for years, most of the text about Under Armour described it as MyFitnessPal's owner, and the correction is a handful of filings and press releases. Nobody outside the model vendors can confirm how a model weighs that, so treat it as a working explanation, not a finding.
Strictly, the first error is an outdated fact rather than an invention. Most people call both "hallucinations". For a shopper reading the answer, the difference doesn't matter.
Why updating your website won't fix a memory answer
ChatGPT and Claude answered these questions from memory, with web access off. Nothing you publish this quarter reaches an answer a model gives from memory. That changes only when the vendor ships a new model.
What you control is what AI finds when it does search, the way Perplexity did here:
- State the current fact on your own site, in plain text. If you sold a product line or changed CEO, say so on a page that loads without JavaScript.
- Make sure AI crawlers can reach that page. Your robots.txt can allow a crawler while your firewall blocks it. Check both.
- Correct the old version wherever you can. Your own old press pages, directory listings, anywhere the before still reads as the present.
- Re-check after every model update. Memory answers only change when the model does.
No markup, file or phrasing has credible evidence that it makes AI favour a page. The levers that hold up are about being reachable and being right. See how LLMs prioritise information about brands for the evidence.
How to run your own AI visibility audit
- Write five questions a buyer asks without your name. "Best [category] for [use]" is the pattern.
- Ask each AI platform and record four things: were you named, were you named first, who was named before you, and did the answer link to your site.
- Ask three questions that do name you: who runs the company, its history, and how it compares with your biggest rival.
- Check every factual claim against a primary source: your filings, press releases, pricing page. Start with anything that changed in the last five years.
- Write down the date and the model. Answers change between model versions, so an undated audit can't be compared with the next one.
By hand, that is an afternoon per brand. 99Visibility runs it across five AI platforms (ChatGPT, Gemini, Perplexity, Google AI Overviews and Claude), tracks your competitors alongside you, and checks what the answers say against the brand facts you enter. Your first audit is free, with no credit card, and covers ChatGPT and Google AI Overviews. See the plans for all five platforms. Inside the dashboard, Agent V reads your audits and your site and drafts the fix, such as page copy for you to paste. It never edits your site itself.
FAQ
How do you check if AI is hallucinating about your brand?
Ask it questions about your brand, then check each factual claim against a primary source: a filing, a press release, your own pricing page. Start with facts that changed, such as prices, leadership, ownership and discontinued products. That is where all three errors in this audit sat.
Which AI hallucinates the most?
Our sample can't rank them. All three confirmed errors in these 67 answers came from Claude, but three errors are too few to compare platforms fairly.
Does Under Armour own MyFitnessPal?
No. Under Armour sold MyFitnessPal to Francisco Partners in December 2020, and sold MapMyFitness, which includes MapMyRun, to Outside Inc. in August 2024. The full timeline is in who owns MyFitnessPal.
See what AI says about your brand. Run a free AI visibility audit: no credit card, your score in under five minutes.
Get your AI visibility score
One number from 0 to 100: how often AI mentions you, how accurately, in what tone, and whether it cites you, scaled by the market weight of the platforms that name you.
Start Free Audit46/100
Named on 3 of 5 platforms× 0.76 coverage
Related Articles
AI Hallucination Examples: 5 Real Cases, With Sources
Five AI hallucination examples, each traced to its primary record: a tribunal ruling against Air Canada, a court order on ChatGPT's fake cases, Google's own post, and the errors we found in Claude's answers about Under Armour and PUMA.
AI Brand Visibility: How to Measure It Without Fooling Yourself
AI brand visibility is four counts, not one number: mentioned, named first, share of brand mentions and links to your site. How to read each per platform and per question type, with a real audit of four sportswear brands.
AI Brand Monitoring: What to Track Every Month, and the Changes Worth an Alert
AI brand monitoring is the same audit, repeated. What to record on each AI platform, why the answers change, the four changes worth an alert, a monthly routine, and how to choose a tool.


