ChatGPT, Claude or Gemini: Which AI Is Actually Best for You?

They can all write an email.

They can all write an email. That does not mean they are the same tool.

A couple of years ago, the question was whether you should be using AI at all.

Now there is another problem.

Which one?

ChatGPT. Claude. Gemini.

Open Instagram for long enough and somebody will tell you one of them has just destroyed the other two. A new model launches. Somebody posts a benchmark. For 48 hours, apparently everything you were using before is obsolete.

Then another one launches.

For most people, this is not particularly useful.

You probably do not need to know which AI scored three percentage points higher on an obscure reasoning test. You need to know which one can help you write the proposal, plan the shoot, research the idea, go through the 40-page document, find something in your files, edit the picture or turn the thought currently sitting in your Notes app into something useful.

I am a photographer, writer and content creator, but a lot of the way I work has always been about systems too. I am interested in the creative part, but also in what makes the creative part easier to actually get done.

So it helps to look at AI in much the same way you would look at putting a team together.

Not: which employee is objectively the best human?

But: what do you need this person to do?

That is a much better way to think about ChatGPT, Claude and Gemini.

Because there is now a huge amount of overlap between them.

All three can write, research, analyse documents, understand images and help you think through an idea.

The interesting part is where they stop overlapping.

This comparison reflects the products as they exist in September 2026. AI moves quickly. The model names will change. The more useful differences are increasingly in the systems being built around them.

First: how many people are actually using them?

A lot.

ChatGPT now has more than one billion weekly active users, according to OpenAI.

Google says the Gemini app has passed one billion monthly active users.

Claude is considerably smaller on the consumer side. Anthropic does not currently publish an equivalent overall monthly-active-user number, but Sensor Tower estimated approximately 245 million monthly users in May 2026.

Those numbers are not directly comparable.

ChatGPT’s figure is weekly. Gemini’s is monthly. Claude’s is a third-party estimate rather than an official Anthropic figure.

So do not use them as a league table.

What they do tell us is that AI assistants have moved a very long way from being experimental tools used by a small group of people interested in technology.

Hundreds of millions of people are now building them into everyday work.

The more interesting question is which one belongs in yours.

01 / ChatGPT

Think: the all-rounder.

If you are choosing one to start with and have nothing else to go on, ChatGPT is probably the easiest place to begin.

Not because it is automatically better at every individual task.

Because it now does so many different kinds of work in one place.

You can use it to research something in the morning, improve an article in the afternoon, analyse a spreadsheet, work through files, create an image, edit a photograph, plan a project and then come back to that same ongoing piece of work later.

That breadth matters.

There is a very big difference between asking “write me an Instagram caption” and having an AI that understands what you make, what you sell, how you write, what you have already published and what you are trying to achieve, and then asking “what should I post today?”

That is where AI starts becoming genuinely useful.

Not when it gives you one impressive answer.

When it starts becoming part of a system.

Where ChatGPT gets particularly interesting: it can do things.

This is one of the biggest differences between simply asking an AI for help and beginning to treat AI like an assistant.

ChatGPT Work can use its own cloud browser to move through supported websites, click buttons, enter information into forms and complete multi-step browser tasks. The desktop version can also work with supported websites and tools directly.

That does not mean you can hand it absolutely anything and disappear. Some websites block automated access, some actions require your approval and some tasks still need you to take over.

But the direction is important.

You can increasingly say: go and do this.

Rather than: tell me how to do this.

For anybody running a business on their own, that difference is enormous.

And for visual creators, ChatGPT can actually work with the image.

You can generate images, upload your own photograph and make changes to it.

Remove something. Add something. Change a background. Extend the frame. Build a graphic. Try another visual direction.

That matters if your work is visual. You do not only need an AI that can describe what could be changed in an image. Sometimes you want it changed.

Use ChatGPT for:

running ongoing projects

brainstorming and developing ideas

planning content

research

analysing files and information

image generation

photo editing

graphics and visual experimentation

organising complicated pieces of work

repeatable workflows

moving between creative work and running the business

tasks where you want AI to take steps rather than only tell you what to do

When ChatGPT might not be right for you

Its greatest advantage can also be unnecessary.

If you simply want somewhere to put a 10,000-word document and spend an afternoon pulling the argument apart, you may prefer the way Claude approaches that kind of deep work.

If practically everything you do already sits inside Gmail, Docs, Drive, Calendar and the wider Google ecosystem, Gemini’s native position inside Google may remove more friction.

And if all you need from AI is one fairly narrow function, you may not need the enormous ecosystem ChatGPT is becoming.

ChatGPT makes the most sense when you want one AI to sit across lots of different parts of your work.

Think less: writer. More: generalist member of staff.

02 / Claude

Think: the deep worker.

Claude is probably the easiest one to misunderstand if you simply compare feature lists.

There is now plenty of crossover with ChatGPT.

Claude can research, work with files, create documents, spreadsheets, presentations and PDFs, analyse data and build things using code.

But Claude makes particular sense when the work itself is the thing you want to sit with.

Give it the report. Give it the article. Give it the strategy. Give it all the notes.

Then talk through them.

For writers in particular, that can be useful.

You can give Claude a substantial amount of your existing material and ask: where have I already made this argument? Which section is weakest? What am I assuming the reader already knows? Does this paragraph actually belong here? Don’t rewrite this yet. Tell me why it isn’t working.

That last one matters.

One of the least interesting ways to use AI is constantly asking it to produce something from nothing.

Sometimes you do not need another writer.

You need an editor.

Claude is very good at feeling like that.

Claude can also make actual files

This is worth saying because the idea that Claude is simply a writing chatbot is outdated.

It can now create and edit Excel spreadsheets, PowerPoint files, Word documents and PDFs, as well as analyse data and build visualisations.

Its Artifacts and visual tools also mean it can make diagrams, charts, interfaces and interactive visuals alongside the conversation.

So this is not a small tool.

But there is one limitation that matters if you work visually.

Claude does not edit your photographs.

Claude can look at an image. It can analyse it. It can tell you what it sees. It can talk about composition, content or what you might change.

But Anthropic explicitly says Claude does not generate photographs or illustrations in the way dedicated image-generation tools do. Its generated visuals are things such as diagrams, charts, HTML and SVG graphics.

That means there is a big difference between uploading a photograph and asking “what could I improve here?” and asking “remove the person behind her, extend the wall on the left and change her coat to black.”

Claude can help you think about the first one.

It is not the tool to choose for the second.

For photographers and visual creators, that is not a small limitation.

Use Claude for:

long documents

editing substantial pieces of writing

interrogating a draft

comparing versions

reports and research

strategy documents

turning messy thinking into clearer structure

creating documents and presentations

coding

deep work where you want to stay very close to the material

When Claude might not be right for you

If you work visually and want your AI subscription to include photo generation and conversational photo editing, Claude is not the first choice.

If you want to move constantly between an article, an image, a business task, browser work and content creation, ChatGPT is currently broader.

And if a lot of what you want AI to do is buried inside Gmail, Drive and the rest of Google, Gemini has an obvious native advantage.

Choose Claude because you want its particular way of working with ideas, writing, documents and code.

Do not choose it as your only AI simply because somebody told you it gives better answers.

03 / Gemini

Think: Google.

This is the easiest advantage to understand.

Google does not just have an AI chatbot.

It already has an enormous part of our digital working lives. Gmail. Google Docs. Drive. Sheets. Calendar. Maps. Chrome. YouTube. Google Photos. Search. Android.

Put Gemini across that ecosystem and you have something potentially very powerful.

Instead of asking “tell me about this company” you can increasingly get closer to “research this company, look at the documents I already have about them and help me prepare for the meeting in my calendar.”

That is where Gemini becomes really interesting.

It is not just about whether it can write a better paragraph. It is about where it lives.

Gemini is also a serious visual AI.

Google has pushed heavily into image, video and multimodal AI.

For a creator, that matters.

Gemini can work with images in a way Claude currently cannot, and Google’s wider ecosystem gives it strong creative potential across image and video creation.

So if you want AI to exist alongside your visual workflow, Gemini deserves serious consideration.

But there is an important distinction around control.

Gemini is not currently the AI that can control everything for you. That is too broad.

To be clear, Gemini can perform actions.

Google says Gemini on Android can automate actions across more than 40 popular apps, and Gemini in Chrome in the UK can already do things such as schedule a Calendar meeting, draft an email in Gmail and work across information in your browser.

So saying Gemini can do nothing would be wrong.

But its action-taking ability is still much more tied to Google’s products, Android, Chrome, specific integrations, your device, your plan and where you live.

That is different from the direction ChatGPT Work is taking, where the proposition is increasingly: give me the task and let me work through the relevant website.

With Gemini, think: Google assistant becoming increasingly agentic.

With ChatGPT, think: general AI workspace increasingly capable of operating across other services.

Those two things may eventually meet in the middle. They are not quite the same thing today.

Use Gemini for:

work centred around Gmail and Google Drive

Google Docs, Sheets and Calendar workflows

finding information across Google’s ecosystem

web research

image generation and editing

visual experimentation

video and Google’s wider creative AI ecosystem

Android users who want AI connected more deeply to their phone

somebody whose digital life already largely belongs to Google

When Gemini might not be right for you

If you do not use Google’s ecosystem much, one of its biggest advantages disappears immediately.

And if what you really want is to hand an AI a fairly arbitrary browser task and have it go off and complete that process across third-party websites, ChatGPT is currently ahead.

Gemini is gaining more action-taking abilities quickly, but they are not all universally available.

That is important because AI demonstrations online are becoming increasingly misleading.

You see somebody in another country, on another plan, with another device, using a new experimental feature and understandably assume Gemini can do that.

It might. Your Gemini might not.

If you are a long-form writer who wants to live inside large documents for hours at a time, Claude is also worth serious consideration.

Gemini’s clearest advantage is not that it universally beats the other two. It is that it is Google’s AI sitting inside Google.

If Google already runs a large part of your life, that is a very big advantage. If it doesn’t, it is much less interesting.

So which one is actually best?

The annoying answer is: it depends.

The useful answer is this.

You want one AI that can do a bit of everything: start with ChatGPT

You create content and also run the business behind it: ChatGPT

You want AI to actually carry out browser-based tasks: ChatGPT

You want to create and edit photographs: ChatGPT or Gemini

You are a photographer and visual AI matters: ChatGPT or Gemini

You write and edit a lot of long-form work: Claude

You spend a lot of time inside substantial documents: Claude

You want AI to challenge and edit your thinking: Claude

Most of your work lives in Gmail, Docs and Drive: Gemini

You use Google for almost everything: Gemini

You want Google’s image, video and Android ecosystem: Gemini

That does not mean ChatGPT cannot write. It can.

It does not mean Claude cannot research. It can.

It does not mean Gemini cannot work outside Google. It can.

This is simply where the differences become meaningful enough to affect which one you pay for.

Do you actually need all three?

I have all three.

Not because more subscriptions make better work, but because they are good at genuinely different things, and the strongest answer often comes from using them in tandem rather than picking a winner.

That does not mean paying three times. The free version of each is enough to sense check: work in the one you pay for, then put the same question to another and see whether it disagrees.

Start with a problem. Then choose the tool.

Where it stops being useful is when the setup becomes the work: six subscriptions, nine browser extensions and a dashboard showing which model currently has a 2% advantage at maths.

If most of your working life already exists inside Google, repeatedly downloading files from Drive to upload them somewhere else might make absolutely no sense.

And if you are a photographer, paying for Claude as your only AI and then discovering it cannot actually make the visual edits you bought AI for is exactly the kind of thing worth knowing before you subscribe.

AI should remove friction. If your AI setup becomes another hobby you need to maintain, something has gone wrong.

You can also use more than one

There is no requirement to become loyal to an AI company.

They are tools.

I use Canon cameras. That does not require me to develop an emotional objection to Nikon.

The same applies here.

You could develop an idea in ChatGPT. Put the 3,000-word article into Claude and ask it to challenge the structure. Use Gemini to find something across your Google ecosystem or experiment with another visual direction.

Then bring the useful parts back into your main workflow.

You can even occasionally give the same problem to two different systems.

Not because AI needs an AI focus group. Because disagreement is useful.

If ChatGPT tells you one thing and Claude tells you something completely different, that can be the point at which you actually need to think.

Which brings us to the part that matters most.

None of them are you

This is the mistake we are going to have to learn our way out of.

AI is getting extremely good.

That does not mean everything it produces should survive.

The paragraph can be beautifully written and still not sound like you.

The content plan can be perfectly organised and still be boring.

The image can be technically extraordinary and mean absolutely nothing.

The research can look convincing and still need checking.

You remain responsible for the taste, the judgement and the decision.

For noticing the thing nobody told the machine to look for.

AI can organise more, research faster, work through information, create things that previously needed entire teams and remove an extraordinary amount of friction between having an idea and making it real.

Good.

I am interested in what happens when more people get access to that kind of power.

But the goal is not to gradually remove yourself from your own work.

It is to remove enough of the unnecessary work that you have more time for the part that actually needs you.

So don’t begin with: which AI is the best?

Begin with: what are you trying to do?

Then choose the one that helps you do more of it.

And if another one does that particular job better next month?

Use that one.

The tools will keep changing.

Knowing what you are trying to make is still your job.