Which AI company should we use? For most businesses it matters less than the marketing suggests. How you adopt it matters more than whose logo is on it. This page is the overview, and each linked guide covers one provider in depth.
Which AI chatbots people use most
ChatGPT leads by a wide margin. The clearest picture comes from Pew Research Center’s February 2026 survey of US adults, which found 49% now use an AI chatbot of some kind. Here’s the ranking, next to the scale the vendors report themselves:
| Provider | Pew: % of US adults using it | Scale, as reported by the vendor |
|---|---|---|
| ChatGPT (OpenAI) | 44% | 900M+ weekly users (OpenAI) |
| Gemini (Google) | 24% | 950M+ monthly app users (Alphabet, Q2 2026 earnings) |
| Copilot (Microsoft) | 17% | bundled across Microsoft 365 |
| Meta AI | 14% | 1B+ monthly users across Meta’s apps (Meta) |
| Grok (xAI) | 8% | integrated into X |
| Claude (Anthropic) | 6% | consumer share understates professional use |
Two things follow from that. Your staff already use these tools, whatever your policy says. And consumer popularity is a poor guide to business fit: Claude’s 6% share tells you nothing about how good it is at the long-document, careful-writing work professional firms do all day.
The six main AI providers
Each card links to that provider’s full business guide. The open-weight models guide covers the models beyond these six names.
OpenAI / ChatGPT
The default general tool: biggest user base, broadest features, strongest ecosystem.
Google / Gemini
The Workspace choice, and the engine behind AI answers in Google Search itself.
Microsoft / Copilot
AI inside Outlook, Word, Excel and Teams, where most UK service-firm work lives.
Anthropic / Claude
The specialist for judgement-heavy writing, long documents and dependable agents.
xAI / Grok
Live access to X and a looser voice: strong for monitoring, evaluate hardest for client work.
Meta / Llama
The assistant inside WhatsApp and Instagram, and the family anchoring open-weight models.
Other providers worth knowing
Three more names have their own guides. Perplexity is the research specialist, with sources attached to every answer. DeepSeek is the price disruptor whose open-weight releases moved the whole market. Mistral is the main European option, the one to look at when your board asks where the AI is built and where the data lives. For evidence rather than positioning, the task-by-task table and our own drafting benchmark show how they compare in real use.
How to choose an AI provider
Three questions settle most of it:
- Where does your work already live? If your firm runs on Microsoft 365, Copilot earns its place by being there already, and the same logic points Google Workspace firms at Gemini. The best model you have to copy and paste into loses to a slightly weaker one that’s already inside the document.
- What is the highest-stakes thing you want AI to touch? For drafting that goes in front of clients or regulators, the differences in quality and reliability are real. Trial the same five real tasks on each candidate and judge the output, not the demo; our own benchmark shows the method.
- What are the data terms? Is your data used for training (on business tiers it shouldn’t be), where is it processed, how long is it kept, and can you get a data processing agreement that satisfies UK GDPR? A vague answer means that provider isn’t ready for that use.
Whatever you pick, keep a way out: don’t build anything that only works with one vendor’s API. The gap between the big providers is the narrowest it has been since 2023, and pricing, models and terms all change every quarter. In my own systems every AI call has a fallback to a second provider and then a third, so an outage or a price change can’t take a system down.
Then write the decision down as an approved-tools list with the one rule that prevents most incidents: client personal data goes only into approved tools on business terms, never into free consumer ones. The UK governance guide covers what sits around that rule.