We tracked Speedo across 3,562 AI responses. Here is what we learned

For five months, we ran the same set of swimwear buying questions through ChatGPT, Claude, and Gemini. The schedule varied from once a day to every hour. We were not testing for a winner. We wanted to see how a single, well-known brand actually shows up inside AI answers, and how that picture shifts over time.

This post is the raw data. No paid placements. No SEO tweaks. No scraped Reddit tricks. Just 3,562 responses from four major language models, captured the same way a customer would see them.

What we tracked

We configured twelve prompts that real swimmers and shoppers actually ask. They fall into three buckets:

• Brand awareness: `Which brand do you consider the leader in competitive swimwear?`, `When you think about swimwear brands, which brands come to mind first?`, `Best swimwear brands for men?`
• Product category: `What are the best options for high-performance swimwear technology?`, `What swimming goggles are comfortable and maintain a good seal?`, `What are the best swimming goggles for indoor and outdoor swimming?`, `What swim gear should I buy as a beginner swimmer?`
• Head-to-head: `arena or speedo to compete`, `which swimsuit is better for competitions?`, `Any brand for training swimsuit that fits a large bust and wider hips?`
• Adjacent: `What are the best options for competitive swim caps?`, `What are the top factors to consider when selecting athlete sponsorship deals?`

Each prompt was sent to ChatGPT, Claude, Gemini 2.5, and Gemini 2.5 Flash. For every response, we recorded:

• Whether the brand was mentioned at all
• The position the brand appeared in (Rank 1, Top 2, Top 3, deeper, or exclusive mention)
• The sentiment of the surrounding text
• The full response, so we could re-check later

The data lives in a single SQLite database per tracked brand. We treat it as a measurement, not a marketing artifact.

The headline number

Across all 3,562 responses, Speedo was mentioned 83.4% of the time. That is a high floor for a brand in a competitive category. Most teams would stop at that number.

We did not. The interesting part is what the 16.6% looks like.

Per-model breakdown

The four models are not interchangeable. Each has a different default posture toward established brands.

Two things stand out:

1. Claude is the most generous to Speedo. It mentions the brand 88% of the time, slightly ahead of ChatGPT. Anthropic's model tends to surface a wider shortlist, which lifts household names.

2. Gemini is the most skeptical. Both Gemini variants sit below 80%. They return smaller, more curated lists. Speedo still appears most of the time, but the gap is real.

The sentiment averages are hard to compare across models because each one uses a different internal scoring scale. The direction matters: Speedo's vibe is positive everywhere, with nowhere near a negative mean.

Where Speedo actually lands

Mention rate is only useful if the brand is high enough in the answer to be seen. We recorded the position for every mention.

Two out of three times Speedo is mentioned, it is the first brand the model names. Another 20% puts it in the top two. Strong visibility profile, but it also means the brand is doing most of its work in the lead slot. If the model rotates the lead, mention rate can drop without anyone noticing why.

Sentiment at a glance

We labeled every response as positive, neutral, negative, or error.

Only 37 responses were negative. Most of those were product critiques (fit, sizing, price, racing suit regulations), not brand attacks. The neutral bucket is the more interesting one. It usually contains `Speedo is a well-known brand that makes X` style framing, where the model acknowledges the brand without describing it. Borderline visibility: the brand is named, but it is not really evaluated.

The single prompt that revealed the gap

Most prompts hovered between 80% and 100% mention rate. One sat far below.

The biggest lesson from this table: Speedo is invisible on men's swimwear brand questions at twice the rate of women's category questions. The gap is not random. Models still default to a women's-first listicle when they default to swimwear at all, and Speedo's brand voice sits more naturally in the competitive lane than the lifestyle lane.

The 0% on sponsorship is a methodology footnote. That prompt is upstream of the category and any brand would score zero.

The five-month trajectory

Visibility is not static. The same brand can move 15 points in a quarter without anyone changing the website.

The April dip is the most interesting data point on the chart. It lines up with a stretch where ChatGPT and Gemini both refreshed their top-of-list defaults for the swimwear category. Speedo did not fall off. It fell to position 2 or 3. Once the model ranks settled, the brand climbed back into the lead slot and the rate recovered.

This is the pattern that matters most for marketing teams. A single mention rate is a snapshot. The trend line is the product.

What the data shows

Strong brands still win in AI search. Eighty-three percent visibility across thousands of prompts is real. The fear that AI would flatten established brands is overblown, at least so far — which lines up with Mordor Intelligence's swimwear market analysis putting the broader swimwear category at USD 28.29 billion in 2025 (projected to reach USD 38.49 billion by 2031), with Speedo's parent Pentland Group and Arena among the five largest players worldwide.

Position matters more than volume. Speedo is mentioned most of the time, but two-thirds of those mentions depend on holding the first slot. If a model rotates the leader, that number moves fast.

Every model has a bias. Claude is more generous to incumbents. Gemini is more selective. ChatGPT sits in the middle. Trying to optimize for one and ignore the rest is a mistake.

The gap is rarely the category. It is the prompt. `Best swimwear brands for men` is not a category problem. The fix is content, not media spend.

How we ran this

A small note on methodology so this post is useful to replicate.

• Each prompt was sent through the official API or first-party interface for the model in question. We did not use search-augmented mode and we did not inject any bias into the prompt string.
• The scraped response was scored by a separate LLM judge, not the model that produced the answer. The judge sees the response and the brand list, and returns a structured JSON record with `brand_mentioned`, `position`, and `sentiment`.
• We stored every response, not just the ones where the brand was mentioned. Negative evidence is the most important data in any tracking study.
• The data spans March 8, 2026 through August 6, 2026. Almost five full months of continuous scanning.

If you want to run the same study for your own brand, the workflow is in LLM Monitor. Setup is under ten minutes.

The scale of this work matters because AI discovery is no longer niche. A Pew Research Center study from July 2025 showed that users clicked a traditional search result in only 8% of visits when an AI summary appeared on Google — meaning the answer you see inside ChatGPT, Claude, or Gemini is increasingly the only brand exposure a buyer will get.

Frequently asked questions