Pressfront

Research

Original research on how AI recommends businesses

We run our own studies on ChatGPT, Gemini, Claude and Google AI to understand how they decide which businesses to name. The data below is ours to cite; attribution to Pressfront is appreciated.

Test · October 2026

Does a press release get you recommended by AI? We tested it.

We sent out a real press release, then asked ChatGPT, Gemini and Claude 48 everyday "best app" questions. The release was used 0 times. Asked about the company by name, AI did read it. What got quoted instead: the competitors' own pages.

0 of 48"best app" answers used it

14 of 23name lookups read it

3AI assistants tested

Read the test →

Study · September 2026

How consistent are AI recommendations? A first look.

We asked ChatGPT and Gemini the same ten "best X for a small business" questions three times each, in one day. About a third of the recommended businesses changed between identical runs, and the churn concentrated in the tail of the list while the top names stayed locked. ChatGPT was the steadier of the two (87% consistent vs Gemini's 52%).

~31%of picks changed on a repeat

56%named in every run

27%named in only one run

Read the study →

Study · September 2026

Link rot in Wikipedia's marketing and PR articles

We audited the citations in 30 of Wikipedia's core marketing, PR and advertising articles. About 70 of their source links are dead, and roughly one in ten of those has no archive backup at all. AI answer engines inherit this citation graph, holes included.

30articles audited

~70dead citations found

~1 in 10dead with no archive

Read the study →

Journalists and researchers: figures on this page are free to cite with a link to pressfront.co. The AI-consistency study's raw data, methodology and analysis code are open on GitHub (60 responses, CC BY 4.0): ai-recommendation-consistency. More studies in progress.

See where your business standsFree AI Visibility Report across ChatGPT, Gemini and Google AI