Do comparison pages help B2B brands get cited by AI?

Do AI assistants look for comparisons when B2B buyers ask?

Yes, at least ChatGPT does, and it often searches the vendors’ own sites too. In our hidden-searches study, we captured the searches assistants ran before answering 80 buyer questions in eight industries, from software to legal services. ChatGPT ran a search that compared options in 33.8% of its answers.

It also ran a search aimed at one option’s own website in 23.8% of answers. Gemini and Claude did neither in our sample: both shares were 0.0%. So the payoff from a comparison page depends on which assistant your buyers use.

Chen and colleagues (opens in a new tab), at the University of Toronto, reached a similar view from AI prompts people shared on Reddit, mostly about consumer purchases. They found decision support dominates: people ask what to buy and how to compare, and expect a shortlist with reasons.

How often do AI engines cite comparison pages?

Not often: ranked “best of” lists are cited far more than head-to-head comparison pages. The vendor Ranqo (opens in a new tab) classified the pages cited across 100-plus brands it tracks on five AI engines. Its author works for the company, and the data come from brands tracked on its platform.

A comparison page, in this scheme, is a page weighing named options against each other, such as one product versus another. Lists that rank many options win far more citations, a pattern we cover in our article on “best of” lists.

Does comparison content shape the answer once a page is cited?

Yes, in the largest public dataset that measures it. Zhang and colleagues (opens in a new tab), independent researchers, analyzed 602 prompts across ChatGPT, Google’s AI and Perplexity. They scored how much each cited page shaped the final answer, through repeated use, early placement and shared wording.

When the researchers looked at the role a citation played, comparisons scored 0.1524, just behind definitions at 0.1531 and well ahead of plain references. Comparison questions had the highest influence of any question type. The authors stress this is a snapshot that shows association, not cause. Whether changing a page’s format alone moves citations is covered in our guide to restructuring existing content.

Do comparisons win the citation against a competing page?

They help a little, but price, topic match and freshness matter far more. Researchers at the software company Sprinklr, Vishwakarma and colleagues (opens in a new tab), ran 252,000 trials across six AI models. Each trial showed a model two product pages that differed in one factor, with brand names removed.

Of 18 factors tested, 11 made a reliable difference in at least four of the six models. Including comparisons with alternatives was one of seven “secondary” factors. Four “gatekeepers” mattered more: matching the topic, stating a price, carrying a recent date and appearing first.

The test used consumer product reviews, not B2B software pages, and only two pages at a time. We cover it in depth in our article on why AI cites a competitor’s page first.

Will AI cite a vendor’s own comparison page?

Sometimes, but independent sources are usually preferred. In Ranqo’s data, only 2.9% of citations pointed at the tracked brand’s own domain, while 75.2% went to other companies in the same space. In other words, vendor pages are cited a lot, but mostly your competitors’ pages, in answers about your category.

Chen and colleagues (opens in a new tab) compared a web-enabled GPT model with Google on 1,000 ranking questions in the US and Canada in 2025. For software products in the US, the AI drew 72.7% of sources from earned media, such as review and comparison sites, and 26.7% from brands. Google’s mix was 43.7% brand, 10.9% social and 45.4% earned.

A 2025 audit of 70 B2B software prompts by Kumar and Palkhouski (opens in a new tab) linked well-built pages to citation. Yet its authors warn that even high-quality pages may not be cited if they sit only on vendor blogs. More on this in our article on whether AI cites your own website.

Do self-serving comparison lists work?

They do get cited, but there is no sign they win the answer. In our study of “best of” lists, 26.0% of cited numbered lists with an identifiable publisher included that publisher. Of those, 92.9% ranked themselves first, and most were software companies writing about their own category.

Such lists were only 1.1% of all citations in our data. Across answers that cited a list, the list’s top pick was named in 47.1% of cases for self-ranking lists and 43.4% for independent ones. In the cleaner comparison, within the same answers, there was no detectable difference.

Do comparison pages help when buyers have not named you?

Mostly no: comparison questions usually already include your brand. In Ranqo’s data, 46% of comparison prompts named the brand being tracked. On average, brands appeared in 74.6% of answers to comparison prompts, against 22.9% for broad discovery questions that named no brand.

That makes a comparison page a tool to defend a shortlist you are already on. It does little to get you onto the shortlist in the first place, which depends more on lists, reviews and coverage elsewhere.

Does accuracy on comparison pages matter?

Yes, because assistants repeat what they find, including stale figures on your own site. In our pricing study, 61.9% of 840 plan prices quoted by four assistants for 45 software products were fully faithful to the pricing page. For 39 of 64 prices that differed, the same figure appeared on another page of the vendor’s own site.

Comparison pages often restate your prices and your rivals’ prices. If they go out of date, an assistant may quote the old figure to a buyer.

What should you do about it?

Publish comparison and definition content where it helps buyers decide, and keep your main effort on third-party coverage.

1. Build pages that answer the comparison, not just praise you. State criteria, prices, specifications and the cases where a rival fits better; these are the parts AI answers reuse.
2. Define your category plainly. Definitions were the most influential citation role in the Zhang dataset, so a clear explainer of what your product category is and does is worth having.
3. Put prices and dates on the page. Both were gatekeepers in the Sprinklr test, and both need regular checks.
4. Keep every page that states a price in sync. Retire or update old comparison pages so assistants cannot pick up stale figures.
5. Invest in independent lists and reviews. Ranked lists and earned media draw most citations for software questions, and they decide who reaches the shortlist.
6. Do not rely on self-ranking lists. They get cited, but in our data their top pick fared no better than an independent list’s.
7. Test on the assistants your buyers use. Only ChatGPT ran comparison searches in our sample, so check results engine by engine.

If you want help deciding which comparison and category pages to build, see our generative engine optimization service.

What does the research not tell us yet?

No study has yet tested whether adding a comparison page to a B2B site raises live AI citations.

• The usage findings are snapshots. The Zhang dataset links comparison content to influence but cannot show that adding it causes more.
• The controlled test is not B2B. The 252,000-trial study used anonymized consumer reviews, two pages at a time.
• The citation shares come from a vendor. Ranqo’s figures come from its own customers’ tracked brands and prompts.
• Comparison pages and lists overlap. Studies classify pages differently, so “comparison” and “best of” shares are not directly comparable.
• Nobody has measured buyer outcomes. No study tracks whether a cited comparison page leads to demos, trials or deals.

Frequently asked questions

Should B2B companies create “X vs Y” comparison pages for AI search?

Yes, if they are honest and specific, but expect modest citation gains. Comparison content shaped answers more in one 602-prompt dataset, yet comparison pages were only 4.5% of cited content in a large vendor dataset.

Do AI assistants cite vendor websites for software comparisons?

Less than independent sources. For US software questions, one 2025 study found AI search drew 26.7% of its sources from brands and 72.7% from earned media.

Do “alternatives to” pages get cited by ChatGPT?

Rarely, on current evidence. Alternatives pages and case studies were 0.7% of cited content in one vendor’s dataset of AI citations.

Does ranking yourself first in a “best of” list work for AI search?

Not clearly. In our study, 92.9% of cited lists that included their publisher put it first, but their top pick was not named more often than an independent list’s.

Are definition pages useful for AI visibility?

They appear to be. Pages with definitions had 57.33% more influence on AI answers in one dataset, and definitions were the most influential role a citation played.

Sources

• Zhang, He and Yao (2026), From Citation Selection to Citation Absorption: A Measurement Framework for Generative Engine Optimization Across AI Search Platforms (opens in a new tab), arXiv:2604.25707.
• Kumar (2026), Generative Engine Optimization at Scale: Measuring Brand Visibility Across AI Search Engines (opens in a new tab), arXiv:2606.20065.
• Chen, Wang, Chen and Koudas (2025), Generative Engine Optimization: How to Dominate AI Search (opens in a new tab), arXiv:2509.08919.
• Vishwakarma, Kumar and Jamidar (2026), What Gets Cited: Competitive GEO in AI Answer Engines (opens in a new tab), arXiv:2605.25517.
• Kumar and Palkhouski (2025), AI Answer Engine Citation Behavior An Empirical Analysis of the GEO16 Framework (opens in a new tab), arXiv:2509.10762.
• Underneath (2026), The hidden searches AI assistants run before they answer
• Underneath (2026), How many “best of” lists cited by AI rank their own brand first?
• Underneath (2026), How faithfully do AI assistants quote software prices?