# Comparison Pages in AI Search: What 1.8 Million Answers Show

We tested comparison pages against 1.8 million AI answers. AI uses them 3x to 11x as often on comparison questions, but they didn't lift brand mentions.

**Published:** October 8, 2026
**Author:** Connor Lahey

---

Comparison pages have become a standard content format for AI search. The usual version goes like this: buyers ask AI "X vs. Y" questions, the engine pulls in comparison pages to answer them, and if you don't publish the comparison yourself, an affiliate will get cited in your place.

We wanted to know whether that's really what's happening, so we decided to test the following questions against Searchable's tracking data:

1. Do AI engines use comparison pages as sources more often when the question is a comparison?
2. Are the comparison pages AI uses mostly written by third parties, more so than other kinds of content (i.e., guides, how-tos, and ranked lists)?
3. When a brand's own comparison page starts getting used by AI engines, does that brand get named more often?

For the first two questions, we used a dataset of 1.8 million AI answers across ChatGPT, Perplexity, Google AI Overviews, Google AI Mode, Gemini, and Copilot. For the third question, we followed 117 brands and compared each one's answers in the four weeks before and after its page started being used, between June and September 2026.

  
  
</KeyTakeaways>

## 1. Comparison questions pull in comparison pages 3x to 11x as often

We split each brand's tracked prompts into two categories: comparison-style prompts (e.g., "Notion vs Coda," "alternatives to Mailchimp," "is X better than Y") and everything else. Then, we measured the proportion of each engine's sources that were comparison pages. Each brand was compared against itself, so categories that are full of comparison content don't inflate the result.

| AI engine | On comparison questions, % of sources that are comparison pages | On all other questions, % of sources that are comparison pages | How much more often comparison pages are used (within each brand) |
|---|---|---|---|
| ChatGPT | 8.4% | 2.1% | 3.2x |
| Perplexity | 13.4% | 1.1% | 10.9x |
| Google AI Overviews | 12.9% | 1.2% | 9.7x |
| Google AI Mode* | 7.0% | 0.8% | 6.2x |
| Gemini* | 18.5% | 1.7% | 10.5x |
| Copilot* | 32.9% | 4.2% | 6.5x |

_*Smaller samples (under 30 brands each), so treat these as directional._

The last column compares each brand with itself and then combines brands, so it comes out lower than dividing the first two columns (8.4% ÷ 2.1% is 4.0x on ChatGPT). The straight division overstates the effect, because brands with many comparison prompts tend to sit in categories where comparison content is common on every prompt.

Every engine shows the same pattern, but the magnitude differs. Perplexity and AI Overviews swing hardest toward comparison pages when the question calls for one. ChatGPT already mixes some comparison content into ordinary answers, so its jump is smaller. (Again, we are treating AI Mode, Gemini, and Copilot data as directional here due to smaller samples.)

**Why it matters:** Comparison pages earn their place when buyers actually ask comparison questions, and do very little outside them. Before you build one, check how many comparison-style prompts exist in your category and how often you show up in them. If you're already doing well on those prompts, or there are few of them, the page likely has less to win.

## 2. Third parties dominate at nearly the same rate across comparison and non-comparison pages

We sorted every comparison page AI engines used in our sample into two groups:

1. Pages on the tracked brand's site or a tracked competitor's site
2. Pages published by third parties such as review sites, affiliates, and publishers (e.g., NerdWallet, Forbes, Consumer Reports, and Tom's Guide).

We also repeated this categorization for guides, how-tos, and ranked lists (combined into a single grouping). The table below shows the proportion of those pages written by third parties.

| AI engine | % of comparison pages written by third parties | % of guides, how-tos, and ranked lists written by third parties |
|---|---|---|
| All engines | 74.5% | 77.2% |
| ChatGPT | 77.4% | 80.3% |
| Perplexity | 70.5% | 77.4% |
| Google AI Overviews | 64.9% | 70.1% |

About three in four comparison pages come from third parties. Every other article format shows nearly the same split, and, if anything, brands wrote slightly more of the comparison pages AI used than of the guides. Other vendors that a brand doesn't track as a competitor count as third parties here (e.g., a HubSpot blog post, when HubSpot isn't on the brand's competitor list), which raises the third-party share for every format by the same rule, so the comparison between formats holds.

**Why it matters:** Third parties supply most of what AI engines read about a category regardless of the format, so publishing your own comparison page won't change who dominates the sources. If you want to show up on comparison questions, the third-party comparison pages already being used as sources are worth as much attention as your own.

## 3. A brand's own comparison page doesn't measurably move its non-branded mentions

So, does publishing a comparison page get your brand named more often? We looked at 117 brands whose own site started being used as a source by AI engines between July and August 2026.

- For 47 brands, the new page was a comparison page.
- For 70 brands, it was a guide or how-to (they are the comparison group since they also had a new page start getting used).

We then compared how often each brand was named in answers to [non-branded prompts](https://www.searchable.com/glossary/branded-vs-unbranded-prompts) the four weeks before and the four weeks after (using the same prompts on the same engines in both periods).

| Group | Number of brands | % of answers naming the brand: 4 weeks before | % of answers naming the brand: 4 weeks after | Change |
|---|---|---|---|---|
| Brand's own comparison page started being used | 47 | 13.5% | 15.6% | +2.1 points |
| Brand's own guide or how-to started being used | 70 | 13.1% | 15.0% | +1.9 points |

Both groups were named a little more often afterward, and by almost the same amount. Comparing brands whose pages first appeared in the same week, the comparison-page group gained 0.2 points more than the guide group, with a plausible range of -1.3 to +1.9 points.

That gap is small enough to be chance, and it stayed close to zero however we cut the data (i.e., later events only, larger brands only, or leaving out answers that cited the brand's own site).

We also ran the same comparison using only the comparison-style prompts these pages are built for, such as "alternatives to [competitor]". For the 85 brands with enough of those prompts, the comparison-page group gained 1.3 points more than the guide group, with a plausible range of -1.8 to +5.0 points. That's no measurable difference either, although with fewer brands and prompts, the range is wide.

That doesn't mean comparison pages are useless. They can still matter in the answers where they appear.

**Why it matters:** Don't expect a comparison page to lift your overall AI visibility on its own. Judge it on the comparison prompts it was built for, and give it longer than a month.

## What to do next

If you're planning comparison content, tie every page to the [comparison prompts you track](https://www.searchable.com/blog/which-prompts-to-track) and measure it on those prompts (not on your overall mention rate). In Searchable, you can group those prompts into their own topic and track your mention rate and the [sources cited](https://www.searchable.com/glossary/ai-citations) on them before and after the page goes live.

![Searchable's Content Types card from a demo account, listing the types of content AI engines cite for a brand's prompts, with Comparison highlighted alongside ranked lists, reviews, guides, and social threads](https://www.searchable.com/blog/comparison-pages-ai-search-content-types.webp)

It's also worth planning for third parties to supply most of what AI sources about you, even if you publish on those topics yourself. For the comparison pages that already show up in your category, that usually means [earned-media work](https://www.searchable.com/blog/aeo-strategy-earned-owned-decision-tree) with the sites that publish them.

## Methodology

### Data

- **Source:** Searchable's [AI visibility tracking](https://www.searchable.com/blog/ai-visibility-tracking). Each tracked prompt is run on each AI engine, and we capture the answer along with the web pages the engine used to build it (i.e., sources).
- **Engines:** ChatGPT, Perplexity (the consumer app), Google AI Overviews, Google AI Mode, Gemini, and Copilot.
- **Findings 1 & 2:** 1.8 million answers with sources from August 24 to September 20, 2026, from a random sample of about 800 tracked brands.
- **Finding 3:** Answers from June 1 to September 27, 2026, for the 117 brands described above.

### Definitions

- **Comparison page:** A page our content classifier labels as a comparison. The classifier labels article-style pages only, so forums, videos, and social posts are almost entirely excluded. In a manual review of 100 randomly sampled pages, 96 compared two or more options; about 15 of those were "best X compared" roundups that sit close to ranked lists.
- **Comparison-style prompt:** A prompt containing "vs," "versus," "compare," "comparison," "alternative(s)," "better than," or "difference between." This rule was fixed before we looked at the results. In a hand check of 100 random matched prompts, at least 90 were genuine comparison questions. The rule misses some phrases, such as "Adyen or Worldpay: which is better?" Accounting for those made the multipliers in Finding 1 slightly larger, so the published numbers are on the conservative side.
- **Brand vs. Third party:** A page counts as brand-owned if it sits on the tracked brand's domain or one of its tracked competitors' domains. Everything else counts as a third party.
- **Named:** The answer mentions the brand by name. Each answer counts once, regardless of how many times the brand appears in it.

### How we measured

- **Finding 1** compares the comparison-page share of sources within each brand (comparison prompts against that brand's other prompts) and pools the result across brands. Ranges come from resampling whole brands 1,000 times.
- **Finding 2** compares the third-party share of comparison pages with the third-party share of guides, how-tos, and ranked lists used in the same answers.
- **Finding 3** compares the before-and-after change in each brand's mention rate between the two groups. It only compares brands whose page first appeared in the same week, and only uses prompts answered in both periods. We then repeated it using only comparison-style prompts that don't name the brand.

---

[Back to Blog](https://www.searchable.com/blog) | [Searchable Homepage](https://www.searchable.com)