Vendors write up to half of lists cited in AI software search

Minuttia has published a study, announced from Estonia on 22 October 2026, on how often Google and AI assistants cite software lists written by the vendors themselves. The agency analysed 18,084 questions put to Google Search, AI Overviews, ChatGPT, Gemini, Claude and Perplexity.

The questions covered 5,983 software brands in 196 categories: 11,525 asked for alternatives to a product, 6,363 asked how two products compare, and 196 asked for the best tools in a category. Researchers then read 311,766 cited pages, which were cited 686,749 times in total, and labelled who published each page.

On alternatives questions, a vendor's own list accounted for 34% of cited listicles on ChatGPT and 47% on Google and Gemini. For comparison questions, the share rose to 56% to 65%. Across all platforms and question types, vendor lists made up 43% of cited listicles. Pages that could not be read were counted as not self-promotional, so these shares are a floor.

The lists followed a similar pattern: 93% highlighted the vendor's product, 87% ranked it first, 90% criticised competitors, and only 16% clearly disclosed that the vendor wrote the page.

The share ranged from 12% to 76% by category. The highest were AI Security Operations Center at 76%, Cloud and FinOps at 74%, AI Evaluation and Observability at 73%, Social Listening at 72% and AI Visibility at 65%. The lowest included Audio Production at 12%, Graphic Design at 16% and Healthcare at 21%. The 20 vendors whose lists were cited most accounted for only 6% of citations.

The platforms gathered sources differently. Of ChatGPT's searches, 71% named exactly one product, and 92% of its site-restricted searches targeted the vendor's own website. On alternatives questions, vendors' sites made up 42% of ChatGPT's citations, against 1% to 6% on the other platforms. Perplexity typically found about 20 pages and cited 9, with citation depending mainly on search position and on features such as a comparison table or a stated selection method.

"Between a third and a half of the lists AI assistants cite when someone asks for software alternatives were written by a vendor in that category, and only 16% of those pages say so," said Zacharias Xiroudakis, managing director at Minuttia. He said each platform picks its sources differently and that the data behind every chart was published so SaaS teams can check the work for their own category.

Every question was asked on all six platforms in August 2026, with one answer per question per platform. The study builds on Minuttia's earlier AI visibility study, uses the same brands and categories, and was conducted with Convex Works. When 2,000 of the questions were put to Perplexity again in September, only 28% of the listed pages had appeared in the August answer. The authors say the findings describe shares of page and site types rather than which individual page wins. The methodology and downloadable chart data are at self-promotional-content-data-study.minuttia.com.