Which brands do AI assistants recommend? 200 answers from ChatGPT, Perplexity, Gemini and Google

AI-generated: an AI agent drafted this post from real runs of a tool I built with it. All the numbers come from 200 answers collected on 26 October 2026, asked from the United States.

More and more buyers ask an AI assistant "what's the best X?" before they open a search engine. So I asked four of them the same 50 buying questions, and counted which brands they named, in which order, and which websites they relied on.

The data comes from AI Visibility Tracker, an Apify Actor I built. It returns one row per question and engine, with the brands named, their order, the sources and the answer.

On 17 of the 41 open questions, all four engines named the same brand first. On 15 more, three of the four did. For most buyer questions, there is one brand the AI layer thinks of first, whatever the assistant.

They split on the vaguest questions. For "best note-taking app", Google put OneNote first, ChatGPT Obsidian, Perplexity OneNote and Gemini Apple Notes. For "most secure VPN", the four answers started with Proton VPN, Mullvad, NordVPN and Mullvad.

HubSpot, 1Password and Wix own their categories: they are named in almost every answer and first in most of them. Other famous brands are named but rarely recommended first:

Add one word to the question and the winner changes:

If you are not the category leader, the questions with a qualifier are the ones you can win.

"Brand websites" are the sites of the brands I tracked.

Across all engines, the most cited website is zapier.com. Its "best apps" roundups appear in 53 of the 198 answers. Review roundups are the pages to be on.

The tracked leader was named in 162 of the 198 answers. In 109 of them, its own website was not among the sources. Engines recommend a brand from what others write about it, so "is my brand named?" and "is my site cited?" are two different metrics.

ChatGPT ran 69 web searches for its 50 answers. 65 of them contained "2026", and many already named brands, for example "best small business CRM software 2026 HubSpot Zoho Pipedrive Freshsales reviews". Pages that carry the current year and a clear comparison have an edge.

One run per question, from the United States, on one day. AI answers vary between runs, so treat small differences with care. "First" means the first tracked brand that appears in the answer text, and "named" means the brand name appears as a whole word.

{ "brand": "HubSpot", "brandDomains": [ "hubspot.com" ], "competitors": [ "Salesforce | salesforce.com", "Pipedrive | pipedrive.com", "Zoho | zoho.com" ], "prompts": [ "best CRM for small business", "best free CRM software", "which CRM is easiest to use" ], "engines": [ "google_ai_overview", "chatgpt", "perplexity", "gemini" ], "country": "US" } Enter fullscreen mode Exit fullscreen mode A check is one question on one engine and costs $0.01; a Perplexity answer counts as 2. This whole study was 250 checks, about $2.60. From an AI agent, connect https://mcp.apify.com/?tools=gaelasao/ai-visibility-tracker and ask the same thing in plain words.

Which category should I test next? Tell me in the comments.