Most ChatGPT value comes from a handful of repeatable workflows, not from clever one-off prompts. Here are the five that cover most people's week. For each one, the cost note uses catalogue prices at the time of writing and assumes GPT-5.4 mini on the API unless stated; check /calculator for your own numbers.
1. Writing: draft, tighten, adapt
Paste your notes or a rough draft, ask for a structured version for a named audience, then iterate: shorter, more direct, add a call to action, produce a LinkedIn variant. Three or four turns is typical.
Cost: trivial on every plan. On the API, a 500-token brief that produces a 1,500-token draft costs about $0.007 on GPT-5.4 mini or about $0.024 on GPT-5.4. Free handles this comfortably; you only need Go or Plus if you write all day and hit limits.
2. Research: search, synthesise, verify
Turn on web search, ask a single well-framed question, and require sources in a table. Then ask for the two claims that matter most and open those sources yourself. For deep multi-source reports, paid plans offer longer-running research modes; they consume more of your usage allowance, so reserve them for questions worth an hour of your own time.
Cost: light research fits Free. Heavy daily research with long reports is where Plus starts to pay for itself. On the API, cost scales with how many source pages you pull into context; a 50,000-token research context on GPT-5.4 costs about $0.13 in input alone before the answer.
3. Coding: explain, fix, generate
Paste the code and the error together, ask for cause and fix, constrain the scope. For new code, describe inputs, outputs and one example, and ask for tests alongside the function. Keep each chat to one file or one problem.
Cost: occasional fixes are fine on Free. Developers who code with ChatGPT all day hit limits quickly and typically move to Plus, or to the API through an editor integration where a session of 20 exchanges with 4,000 tokens of code context each runs well under a dollar on a mini model. See /tutorials/claude-101 for the main alternative developers compare against.
4. Data: clean, analyse, chart
Upload a CSV or spreadsheet and ask plain-language questions: what are the top ten by revenue, which rows have missing dates, chart monthly totals. ChatGPT can run code on the file, which is more reliable than asking it to do arithmetic in prose. Ask to see the code it ran when a number matters.
Cost: file analysis is available on Free with limits and is one of the features that meaningfully improves with a paid plan (bigger files, more runs). On the API you would build this yourself, so for non-developers the subscription is the practical route.
5. Planning: options, risks, schedules
Describe the goal, the constraints and the deadline, and ask for three plans with trade-offs before asking for the detailed version of one. Then ask what could go wrong. Then ask for a week-by-week schedule as a table you can paste into a spreadsheet.
Cost: negligible. This workflow is pure text and fits every plan. It is also where a Project (paid plans) earns its keep, because you can keep the plan, its files and your instructions in one place.
Knowledge check
Which workflow benefits most from upgrading from Free to a paid plan?