Gemini is Google's AI chatbot, and for many people it is already sitting inside apps they use every day.
That familiarity raises its own questions. What does Gemini actually do? Is it free? What does it know about the person using it? And how does it compare to ChatGPT?
This guide covers the basics in plain language: what Gemini actually is, what it genuinely helps with, what the free and paid plans offer, what it remembers by default, and where its limits are. No jargon, no hype, just a clear starting point.
What Gemini actually is
Gemini is really two things, and beginners often mix them up.
The first is the Gemini app: the chatbot people actually talk to, available at gemini.google.com, in mobile apps, as the assistant on Android, and inside Google apps like Gmail and Docs. The second is the family of AI models built by Google DeepMind that powers it.
People use the app. The app runs on the models. Which model answers a question affects the quality of the response. As of mid-2026, the everyday default in the app is Gemini 3.5 Flash, with the more capable Gemini 3.1 Pro selectable from a dropdown for harder tasks like advanced maths and coding, according to the official release notes.
Like other chatbots, Gemini is built on a large language model. It generates responses based on patterns learned during training, which means it can be genuinely helpful and confidently wrong in the same conversation.
What Gemini can help with
Gemini handles the same core tasks as other AI chatbots: drafting text, explaining unfamiliar topics, summarising documents, brainstorming ideas, and answering general questions.
Its distinctive strength is different. Gemini connects directly to Google's own apps, including Gmail, Calendar, Drive, Photos, YouTube, and Maps, and most of these connections are enabled by default once someone is signed in (Tech Nerdiness; Google Workspace Help). That means it can do things a standalone chatbot cannot, such as finding details buried in an inbox or pulling from files in Drive when asked.
Examples include:
- summarising a long email thread in Gmail
- checking Calendar and helping plan around existing commitments
- pulling information from Docs, Sheets, and PDFs stored in Drive
- getting directions and place suggestions through Maps
- answering questions about a YouTube video
Gemini's Deep Research feature takes this further. It can compile research reports that draw on Gmail, Drive, and Google Chat alongside web sources, per Google's announcement.
There is also Gemini Live, a real-time voice conversation mode, and image generation and editing, both available on the free tier (gemini.google/subscriptions).
For someone already living in Google's ecosystem, this integration is the main reason to pick Gemini. For someone who is not, most of that advantage disappears.
Free vs paid: What each offers
One point of confusion first: the paid plans are not sold as "Gemini" subscriptions. They are sold as "Google AI" plans, and each one also upgrades the cloud storage on the same Google account.
Google changes pricing without much notice, so it is worth checking gemini.google/subscriptions before making any decision. As of mid-2026, the UK prices on that page are:
- Free (£0): the Gemini app with Gemini 3.5 Flash, some access to Gemini 3.1 Pro, image generation and editing, limited Deep Research, Gemini Live, and 15 GB of storage
- Google AI Plus (£4.49/month): 2x higher usage limits, video generation, family sharing for up to 5 people, and 400 GB of storage
- Google AI Pro (£18.99/month): 4x higher usage limits, fuller access to Gemini 3.1 Pro and Deep Research, YouTube Premium Lite in select countries, and 5 TB of storage
- Google AI Ultra (from £79.99/month, top tier £189.99/month): first access to advanced features, the highest usage limits, YouTube Premium, and 20 TB of storage. In the US, Google's announcement puts these tiers at around $100 and $200 per month (blog.google).
Ultra got noticeably cheaper in May 2026. Google introduced the new entry tier at I/O 2026 on 19 May 2026 and cut the top tier from around $250 to around $200, a change Engadget describes as a 60% decrease at the entry level. Even so, Ultra is a professional tier, not a consumer one.
Usage limits also changed at the same time. Instead of daily message counts, limits are now based on computing allocation that resets every five hours until a weekly maximum is reached (blog.google). Google does not publish exact message counts per tier.
For most beginners, the free tier is a full product, not a demo. The sensible approach is to start free and only consider AI Pro if usage limits become a regular obstacle and the bundled storage would actually get used.
What Gemini remembers (and what it doesn't)
This is where Gemini differs most from its rivals, and it deserves plain language: Google's defaults favour Google, not privacy.
Two settings are on by default, per the Gemini Apps Privacy Hub and Google's privacy controls announcement:
- Personal context. Gemini can reference past chats to tailor its answers. Turning it off stops the tailoring but does not delete history.
- Keep Activity. Chats are saved to the Google account and used to improve Google's services, including training its AI models. A sample of chats may be read by human reviewers, and reviewed chats are kept for up to three years, disconnected from the account. Saved activity auto-deletes after 18 months by default.
Google's own warning on that page is worth quoting: "Please don't enter confidential information that you wouldn't want a reviewer to see."
There are controls. Turning Keep Activity off limits retention to 72 hours and stops chats being used for training. Temporary Chats offer a one-off conversation that stays out of history, personalisation, and training, and is discarded after 72 hours.
The practical takeaway: leaving the defaults alone is fine for everyday questions, but passwords, financial details, and confidential work material should never go into a chat. Anyone privacy-conscious should turn Keep Activity off and use Temporary Chats for sensitive one-offs.
Limitations and things to watch for
Google is unusually direct about this. Its own support documentation states that Gemini "can hallucinate and present inaccurate information as factual" and specifically flags answers about people as an area where responses "may be inaccurate or inappropriate".
A hallucinated answer reads exactly like a correct one. Anything that matters, numbers, names, dates, claims, is worth checking independently before relying on it. Treat every answer as a starting point rather than a finished answer.
Gemini includes a double-check feature that uses Google Search to highlight statements: green where Search found similar content, orange where it found something different or nothing at all. One caveat matters here. A green highlight means Search found something similar, not that the statement is true.
Real-time information is not guaranteed either. Gemini can pull live data through Search and connected apps, but a plain chat answer may still rest on training data. Prices, news, and schedules all need a check.
Finally, the privacy defaults covered above are a limitation in their own right. Rival tools have their own data practices, but Gemini's defaults are more data-hungry than most, and knowing that upfront is part of using it sensibly.
How Gemini compares to other AI tools
At around £20 per month, Google AI Pro sits directly against ChatGPT Plus and Claude Pro. The core experience is similar across all three: type a request in plain language, get a generated response back.
The differences are at the edges. Gemini's edge is Google app integration and the bundled storage. ChatGPT's is its broad feature set and large ecosystem, covered in the guide to everything ChatGPT.
For most beginners, the practical gap between the big tools is smaller than it seems. The better approach is picking whichever fits existing habits, learning to use it well, and switching later only if a specific need is not being met.
Final thoughts
Gemini is a capable everyday tool with one genuine differentiator: it lives inside the Google apps millions of people already use. For anyone whose email, files, and calendar are already with Google, that integration is worth more than any model comparison chart.
The honest trade-off is data. Gemini's defaults save chats, use them for training, and allow a sample to be read by human reviewers. None of that is hidden, and all of it can be changed, but it should be a deliberate choice rather than something discovered later.
Beyond that, the usual rules apply. Gemini can be wrong while sounding certain, and it works best as a helper, not the final authority. Start free, use it on real tasks, keep sensitive information out of it, and check anything that matters.
