Sometime around mid-2024, I had eleven AI subscriptions running simultaneously. Eleven. I know this because my wife spotted them on the bank statement and, very calmly, asked whether I was funding a startup she didn’t know about. I wasn’t, obviously. I was just doing what every tech-curious person was doing: signing up for every shiny new AI tool that promised to revolutionise my workflow, my creativity, or my meal planning. Two years on, the dust has settled. Most of those tools are gone. A handful survived. The reasons why might surprise you.
The truth is, the AI tool landscape in 2024 was a bit like the early days of streaming. Everyone launched a platform, everyone wanted your monthly subscription, and the average person ended up paying for six services while only actually watching Netflix. I fell into the same trap with AI tools. So here’s my honest, no-hype audit of what I actually use every single week in 2026, what I’ve quietly cancelled, and what I learned about separating genuine utility from the dopamine hit of trying something new.
I should say upfront: I’m not sponsored by any of these companies. I pay for everything myself, which means value matters. When you’ve got three kids, a mortgage, and a Tesla to charge, every subscription has to earn its place.
The Survivors: Tools That Earned Their Keep
ChatGPT Plus is the tool I open most often, and it’s not even close. I use it for drafting articles, brainstorming ideas, summarising long documents, and occasionally helping my youngest get his head around science homework without just handing him the answer. The introduction of GPT-5 on the Plus plan made a noticeable jump in quality, particularly for longer, more nuanced writing tasks. At £24 a month (including VAT), it’s not cheap, but I genuinely use it every day. The memory feature means it knows the context of my ongoing projects, which saves me re-explaining things constantly. The deep research mode has replaced a huge chunk of the Googling I used to do for articles. If I could only keep one AI subscription, this would be it.
Claude Pro is my second brain, and it excels at things ChatGPT sometimes fumbles. Anthropic’s Claude Sonnet 4.6 is brilliant for structured analysis, long-form editing, and anything that requires careful, methodical reasoning. I use it specifically when I’m writing comparison pieces or technical guides because it’s exceptionally good at maintaining consistency across a 2,000-word article. The Extended Thinking feature on the Pro plan is genuinely useful for complex topics. At $20 a month (roughly £16), it’s worth it alongside ChatGPT because they complement each other rather than overlap. Claude handles precision. ChatGPT handles versatility.
Perplexity Pro rounds out my core three, and it fills a completely different role. I use it as my research engine. Rather than locking you into one model, Perplexity routes your queries across GPT-5.2, Claude Sonnet 4.5, and Gemini 3 Pro, pulling the best answer from whichever model handles it best. For fact-checking, sourcing claims, and getting quick answers with citations, nothing else comes close. At $20 a month, it’s replaced my need for several standalone research tools. When I’m prepping for CES coverage or writing a buying guide, Perplexity is the first tab I open.
The Fallen: Tools I Tried and Dropped
Google Gemini Advanced (now Google AI Pro) was the subscription I wanted to love. Google’s 2-million-token context window is genuinely impressive on paper, and I thought the deep integration with Google Workspace would be a game-changer. In practice, I found myself never reaching for it. The responses felt slightly more generic than Claude or ChatGPT, and since I run a Mac Mini M4 as my main home machine rather than living entirely in the Google ecosystem, the Workspace integration didn’t pull its weight. At $19.99 a month, it wasn’t bad value. It just never became a habit. Cancelled in late 2025.
Midjourney was the heartbreaker. I used it obsessively through 2024, generating everything from blog header images to silly family Christmas card concepts. The image quality is genuinely stunning. But the pricing structure became hard to justify for my use case. The Basic plan felt too limited, and jumping to Pro at $60 a month was steep for someone who generates maybe 15 to 20 images a week. When ChatGPT’s built-in image generation improved dramatically with GPT-5, I found myself using that instead for 90% of what I needed. Midjourney produces better results if you’re a digital artist. For a dad who needs a decent blog image and the occasional custom print for the 3D printer, ChatGPT’s image tools are good enough. Sometimes “good enough” wins.
Microsoft Copilot Pro lasted exactly two months. The promise was irresistible: AI baked directly into Word, Excel, and PowerPoint. The reality was clunky. It felt like an add-on rather than a native feature, and I kept finding myself copying text into ChatGPT anyway because the results were better. At $20 a month on top of the Microsoft 365 subscription I already pay for, it felt like I was paying extra for a feature that should have been included. I know Microsoft has improved it since, but first impressions matter, and mine was “this isn’t ready yet.”
Grok via SuperGrok was a curiosity subscription more than anything. I wanted to see what xAI was building, and the $30 a month price tag was already a stretch. It has some interesting capabilities, particularly around real-time information from X (formerly Twitter), but I couldn’t find a weekly use case that justified the cost. If you live on X and need AI analysis of trending topics, it might work for you. I don’t, so it didn’t. Cancelled after three months.
What Actually Matters: My Weekly AI Stack at a Glance
| Tool | Monthly Cost | Primary Use | Verdict |
|---|---|---|---|
| ChatGPT Plus | £24/month | Writing, research, daily tasks, image generation | Essential. Use it daily. |
| Claude Pro | ~£16/month ($20) | Long-form editing, structured analysis, technical writing | Essential. Complements ChatGPT perfectly. |
| Perplexity Pro | ~£16/month ($20) | Research, fact-checking, sourced answers | Essential. Best research tool available. |
| Google AI Pro | $19.99/month | General assistant, Workspace integration | Dropped. Never became a habit. |
| Midjourney | From $10/month | Image generation | Dropped. ChatGPT’s images are good enough now. |
| Copilot Pro | $20/month | Office integration | Dropped. Felt unfinished. |
| Grok (SuperGrok) | $30/month | Real-time social analysis | Dropped. No regular use case. |
| Cursor Pro | $10/month | Code assistance for web projects | Kept for specific tasks. Not weekly. |
My total monthly AI spend has gone from well over £150 at the peak of my “try everything” phase down to roughly £56. That’s a subscription I can justify to myself and, more importantly, explain with a straight face at the dinner table.
Hype Cycle Check: Where AI Tools Really Stand
LIKELY TO LAST: ChatGPT and Claude have established themselves as genuinely indispensable tools for knowledge workers, creators, and curious parents alike. Perplexity’s multi-model approach is clever and future-proof. These three are not going anywhere.
WATCH CLOSELY: Google AI Pro could become essential if Google tightens the integration with Android, Chrome, and its wider ecosystem. The 2-million-token context window is a real technical advantage that nobody else matches yet. Cursor and Claude Code are reshaping how developers work, with Claude Code alone reportedly hitting around $2.5 billion in annualised revenue. If you build websites or tinker with code, these are worth monitoring.
VAPOURWARE RISK: The flood of niche AI tools that launched in 2024, the “AI meal planners,” the “AI wardrobe organisers,” the “AI homework tutors” with flashy landing pages and thin functionality. Most are gone. The ones that survive will be absorbed into the platforms above rather than standing alone. Be very cautious about subscribing to single-purpose AI tools that a general assistant can already handle.
What This Means for CES 2027
Having attended CES roughly eight times over the years, I can tell you that the show floor is a lagging indicator, not a leading one. The AI tools dominating CES booths are usually 12 to 18 months behind what’s actually useful in daily life. That said, I expect CES 2027 to focus heavily on AI hardware integration, specifically devices that run AI models locally rather than relying on cloud subscriptions. Think dedicated AI chips in everything from laptops to home routers. The real story at CES 2027 won’t be “look at this new chatbot.” It’ll be “look at what runs without an internet connection.” For families, that shift matters enormously. Local AI means better privacy, lower latency, and no monthly fees. Having already experimented with self-hosted LLMs using Ollama and LM Studio on my home setup, I can confirm that local AI is improving fast. It’s not quite there for everyday family use yet, but it’s getting closer every quarter.
Recommended on Amazon
These are affiliate links — if you buy through them, Tech Dads Life earns a small commission at no extra cost to you.
What to Watch Next
OpenAI’s ChatGPT Go tier is fascinating. At $8 a month in the US, with no limits on GPT-5.2 Instant, it could become the default recommendation for families who want AI access without the full Plus price tag. I’m watching to see when it launches in the UK and what the GBP pricing looks like.
Claude’s Google Workspace integration is something I plan to revisit. Being able to connect Claude directly to Google Docs and Gmail could change my workflow significantly, especially for collaborative projects. If it works as smoothly as Anthropic claims, I might find myself restructuring my entire stack.
Mistral’s Le Chat Pro at $14.99 a month is the underdog I’m most curious about. It’s cheaper than every major competitor, and Mistral’s open-source roots give it credibility in the privacy-conscious community. I haven’t tested it enough yet to recommend it, but it’s on my radar for a proper trial this summer.
Local AI development remains my passion project. Running models through Ollama on my Mac Mini M4 has gone from a fun experiment to something genuinely practical for certain tasks. The models are getting smaller and smarter simultaneously, and I suspect that by CES 2027, we’ll see consumer hardware specifically designed to make this accessible to non-technical families.
The honest truth about AI tools in 2026 is that less is more. Three good subscriptions beat eleven mediocre ones. The tools that survived in my workflow are the ones that saved me time without creating new complexity, and that’s the bar every new AI product should have to clear. If you want to stay up to date with my ongoing experiments, real-world reviews, and the occasional dad joke about robots taking over the school run, sign up for the Tech Dads Life newsletter at https://techdadslife.beehiiv.com/. No spam, no hype, just honest tech for real families.
