How Many AI Subscriptions Do You Really Need?
ChatGPT. Claude. Gemini. Copilot. Grok. Perplexity.
If you follow artificial intelligence closely, it can sometimes feel as though every useful new feature arrives with another monthly subscription attached.
Individually, many of these services can be relatively affordable. But start combining two, three or four of them and the cost can quickly become noticeable.
The obvious question is: how many AI subscriptions do you actually need?
The answer is probably not determined by which AI comes top of a benchmark table. It depends much more on what you actually use AI for.
Start with the job, not the AI
It is easy to approach AI tools backwards.
A new model launches. People talk about how impressive it is. You try it. Perhaps you subscribe. A few months later, you have several AI services doing broadly similar things.
A better starting point is to ask:
What am I actually paying this AI to help me do?
For one person, that could be writing, brainstorming and analysing documents.
Someone else may mainly want help with software development.
A jobseeker might use AI for CV preparation, interview practice and researching employers.
A small business might want help analysing spreadsheets, preparing content or improving everyday office work.
And someone carrying out regular research may value an AI service designed specifically around finding and working with information from the web.
Those needs do not necessarily require the same product.
They also do not necessarily require five subscriptions.
There is more overlap than there used to be
One reason the decision is getting harder is that the major AI platforms increasingly overlap.
ChatGPT can work with documents, images, research tasks, coding and voice as well as ordinary conversation. ChatGPT Plus is listed at US$20 per month and provides higher usage limits plus access to features including file analysis, image generation and Deep Research where available.
Claude has its own free and paid tiers. Claude Pro is listed at US$20 when billed monthly, before applicable tax, and includes increased usage, Projects, Research and access to more Claude models. Some features, such as app connections, are also available on its free tier.
Google's paid AI plans combine its Gemini models with features across Google's wider ecosystem, including Gmail and Docs.
Microsoft now includes Copilot features inside Microsoft 365 Personal in the UK, alongside Word, Excel, PowerPoint, Outlook and OneDrive. That means somebody already paying for Microsoft 365 may already have access to useful AI features without buying another standalone AI subscription.
Grok also has free and paid options. SuperGrok offers higher limits and additional capabilities, while Grok's wider toolkit includes search, coding, voice, image and video generation.
Then there is Perplexity, which complicates things further.
Unlike a service centred mainly around one company's models, Perplexity Pro is strongly focused on search and research and includes access to models from several providers. That means somebody who occasionally wants to use different AI model families may not necessarily need to subscribe to every original provider separately.
Plan details checked on 19 September 2026. Prices, limits and features can change; currency, tax, region and billing method can affect the price you see at checkout.
The important point is not that these products are identical. They are not.
It is that the areas of overlap are becoming large enough that paying for everything “just in case” may no longer make much sense.
Are you paying for the model or the product?
This is an increasingly useful distinction.
You may want access to a particular AI model, but that does not automatically mean you need the company's full subscription product.
For example, I can use Claude models through Perplexity as part of my own AI setup. That does not give me the complete Claude Pro experience, including all of Anthropic's native features, but it means I do not currently need a separate Claude subscription simply to access Claude occasionally.
The reverse can also be true.
If you work heavily inside Microsoft's applications, Copilot's integration with Word, Excel and Outlook might be more valuable to you than the underlying model comparison.
Someone who spends much of their working day in Google's ecosystem may reach a similar conclusion about Gemini.
And someone who relies heavily on ChatGPT's wider collection of tools might reasonably choose that as their main AI workspace.
In other words, sometimes you are not really paying for an AI model at all.
You are paying for the environment built around it.
If that environment can also act on your behalf, permissions matter as well as features. This is explored in AI Agents Can Act for You — But What Should You Let Them Do?.
Try the subscription audit
Before adding another AI subscription, ask six questions.
What specific job will this tool perform?
If the answer is simply “it is another good AI”, that may not be enough.
Can something I already pay for do the same job?
There is little value in adding another £15 or £20 every month to perform a task your existing AI handles perfectly well.
Would the free tier be enough?
Occasional use does not always justify a subscription.
Will I use the additional feature regularly?
One impressive demonstration is not the same as a recurring need.
Does it integrate with software I already use?
Integration can matter more than small differences in model performance.
What would I actually lose if I cancelled it?
This may be the most revealing question of all.
If you cannot immediately identify what would become harder, slower or impossible, you may have found a subscription you no longer need.
One main AI may be enough
For many people, a sensible starting point is one general-purpose AI assistant.
Use it properly. Learn what it can and cannot do. Explore the tools already included in the subscription before automatically paying for another service.
Then add another product only when it solves a recurring problem your current setup does not handle well enough.
Mercia AI's PDF guides, including the AI-Powered Holiday Planner, offer examples of practical tasks you can try with the tools you already have.
That might mean combining a general AI assistant with a specialist research service.
For a developer, it could mean adding coding tools.
For someone creating a lot of visual content, image or video generation might justify a separate service.
For another user, the free versions of several tools alongside one paid subscription may be entirely adequate.
And some people may discover that they do not need a paid AI subscription at all.
There is no universally correct number.
More AI subscriptions do not automatically mean more value
Artificial intelligence is moving quickly, and curiosity is part of the appeal. Trying different systems can be useful, particularly when their capabilities are changing so frequently.
But experimenting with an AI and paying for it every month are different decisions.
The goal should not be to assemble the largest collection of AI subscriptions.
It should be to build the smallest combination of tools that reliably helps you do what you actually need to do.
Before buying the next one, ask a simple question:
What will this subscription let me do that I cannot already do well enough?
If there is a clear answer, it may be money well spent.
If there isn't, the best AI subscription may be the one you decide not to buy.
If you would like to explore how AI could fit your work or career, Mercia AI's free, 15-minute AI Discovery Call is an introductory conversation to understand your needs and whether our services are a suitable next step. It is not a full consultation or subscription audit. Please request a time at least 48 hours ahead; service eligibility restrictions apply.
