In 2025-2026, the AI tools market exploded into hundreds of products promising to revolutionize your workflow. We tested 50+ as part of our editorial work. Most were forgettable. Five have stayed in our personal and editorial workflow for over six months and are genuinely worth paying for. This article is about those five.
The criteria for inclusion: we use it weekly, it solves a problem we previously solved by hand or with multiple tools, and the price is justified by the time saved.
1. Claude (Anthropic) for writing, reasoning, and code review
What it is: an AI assistant by Anthropic, accessible via web (claude.ai), iOS/Android apps, API, and integrations.
What we use it for:
- Editing long-form articles for clarity and tone
- Reasoning through complex problems where multiple variables interact
- Code review and debugging assistance (Python, JavaScript, Bash)
- Reading dense PDFs (legal documents, research papers, financial reports) and extracting key points
- Translation (English-Italian, English-Spanish, with cultural register awareness)
Why Claude specifically: Claude 4.7 (current as of May 2026) is the most capable model for long-context reasoning and produces the most natural prose of any LLM we have tested. ChatGPT-5 is competitive but feels more “marketing-tone” in default output. Claude tends to be more thoughtful and less prone to confident hallucination.
Pricing: Free tier with usage limits / Claude Pro $20/month for higher limits and access to all models including Opus / Claude Max for power users.
Limitations: cannot browse the web by default (Claude Pro has limited search). Knowledge cutoff date means recent events require manual context. Cannot generate images yet.
2. Perplexity for research with citations
What it is: a search engine that uses AI to synthesize answers from current web sources with inline citations.
What we use it for:
- Research questions where citations matter (“what is the current EU regulation on X”)
- Verifying claims I read elsewhere (“did this paper actually say that”)
- Finding niche tools or services I do not know the name of (“what is the tool for X”)
- Replacing Google for any query where I want a researched answer instead of a list of links
Why Perplexity specifically: the citation linking is the single most valuable feature in AI search. Every claim is sourced to a URL you can verify. This is the antidote to LLM hallucination: you can check whether the answer is actually supported by the sources.
Pricing: Free tier (limited Pro searches) / Perplexity Pro $20/month for unlimited Pro searches with deep web access.
Limitations: the Pro tier is necessary for serious use (free is too rate-limited). Quality varies by query type (excellent for factual research, less reliable for opinion or creative tasks).
3. Raycast AI Commands for in-line text transformation
What it is: Raycast (macOS productivity tool) with Pro Plus tier ($16/month) adds system-wide AI commands.
What we use it for:
- Translate any highlighted text in any app with Cmd+K (Italian, Spanish, French)
- Summarize long emails into 3 bullet points
- Convert casual writing to formal (and vice versa)
- Fix grammar in any text field
- Rewrite a sentence for clarity
- Generate code snippets without opening an editor
Why Raycast AI Commands specifically: the in-line, no-app-switching workflow is the killer feature. You do not break flow to open ChatGPT, paste, copy back. You highlight text in Slack, Cmd+K, “Translate to Italian”, and the result replaces the selection. After a week, you cannot remember life before this.
Pricing: Raycast free / Pro $8/month / Pro Plus $16/month (this is the tier you want for AI).
Limitations: macOS only as of 2026. Windows version in beta. The $16/month is a meaningful cost on top of other subscriptions.
4. Notion AI for knowledge base augmentation
What it is: AI features built into Notion, the document/database platform.
What we use it for:
- Querying your own notes (“summarize what I have written about X”)
- Drafting recurring documents (weekly review, project briefs) from templates
- Translating Notion pages between languages
- Auto-filling repetitive database fields
Why Notion AI specifically: the AI works within your knowledge base, not from generic internet knowledge. When you ask “what did I think about X last month”, it queries your notes, not Wikipedia. This is genuinely different from using ChatGPT separately.
Pricing: Notion AI add-on $10/month per user on top of your Notion plan (Free / Plus $10/user / Business $20/user).
Limitations: only useful if you already use Notion for serious knowledge management. If you do not have substantial content in Notion, the AI features add limited value.
5. ChatGPT (OpenAI) for image generation and voice
What it is: OpenAI’s flagship AI product with text, image, and voice capabilities.
What we use it for:
- Image generation (DALL-E 3 integration, especially for blog post headers)
- Voice conversations during walks (“Advanced Voice Mode”)
- Code Interpreter for data analysis with file uploads
- Custom GPTs for repeated workflows
Why ChatGPT specifically: the multimodal integration is the strongest in the market. You can drop a PDF, an image, a CSV, and a screenshot into the same conversation, and ChatGPT handles all of them. Claude is text-stronger but ChatGPT is multimodal-stronger.
Pricing: Free tier (GPT-3.5 with limits) / ChatGPT Plus $20/month for GPT-4o, image generation, voice / Team / Enterprise.
Limitations: the model output quality has declined in our subjective testing compared to early 2024 (likely due to safety adjustments). For pure text reasoning, Claude is now the better choice in our workflow.
What we removed from our workflow
Tools we tested but no longer use:
- Jasper, Copy.ai: generic AI writing tools. Outclassed by direct ChatGPT/Claude usage at lower cost.
- Midjourney: still excellent for art, but for our use case (blog headers) DALL-E 3 in ChatGPT is sufficient and integrated.
- Otter.ai, Fireflies: meeting transcription. We use macOS built-in dictation + Raycast cleanup, which costs $0.
- Custom fine-tuned models: tried setting up. Cost and complexity not justified for our use cases.
- Local Llama installations: technically interesting, slower than cloud APIs, worse output. Returns to “use the cloud” for most consumer scenarios.
Combined monthly cost
For the full five-tool workflow:
- Claude Pro: $20
- Perplexity Pro: $20
- Raycast Pro Plus: $16
- Notion AI: $10
- ChatGPT Plus: $20
Total: $86/month.
This is significant. Cost-saving alternatives:
- Skip ChatGPT Plus if you do not need image generation ($66/month)
- Skip Notion AI if you do not use Notion ($76/month)
- Use only Claude + Perplexity ($40/month, the minimum recommended)
For knowledge workers, $40-86/month replaces 10-15 hours/week of manual work. The ROI is favorable.
How to start
If you are not paying for any AI tool:
- Start with Claude Pro ($20/month). Use it daily for 2 weeks.
- Add Perplexity Pro ($40/month total) for research-heavy work.
- If on Mac, add Raycast Pro Plus ($56/month total) for system-wide AI integration.
- Evaluate further additions based on your specific workflow gaps.
Do not start by buying all five. Start with one, integrate it into your daily work, then add the next. The risk of buying a $200/month stack is that you do not use any of them deeply because you are too busy switching between them.
The principle: pick the AI tool that solves your most acute current bottleneck. Master it. Then move on.
That is the From Noob to Ninja path in AI tools.