Historical perspective: information available by 2025-04-28. Chronology corrected on September 12, 2026.
ChatGPT has grown beyond a text box, but a longer feature list does not necessarily make it easier to use. Start with the job you want to finish: understand a document, draft something, investigate a question or examine some data. Then choose the relevant tool and check the result.
Writing, search and visual work
Ordinary conversation supports drafting, rewriting, explanation and translation. Search adds current web information; deep research is intended for a more involved investigation across sources. Uploaded images can be discussed, and image generation can create or edit visuals. Voice provides spoken conversation, while Canvas offers a workspace for developing writing or code. Projects organise related conversations and files; memory can personalise later conversations when enabled. Some accounts also offer scheduled tasks. Availability depends on the plan and settings: see OpenAI’s capabilities overview.
For a writing task, try a concrete brief: “Rewrite this announcement for a reader who has never used our service. Keep the date and price unchanged, use fewer than 150 words, and flag anything unclear.” Compare the output with the original before using it. For research, open the cited pages and check whether they support the particular claim being made, rather than treating the presence of a link as verification.
Working with files and numbers
ChatGPT can analyse supported documents and structured files, including spreadsheets, and use code to calculate or plot results. File limits vary with the environment and plan. Clean column names and a consistent row structure help; scans and numbers embedded in images can be harder to interpret accurately. Review the calculation method and assumptions, not just the finished chart. OpenAI’s data analysis documentation explains the supported workflows and limitations.
For example, with a monthly sales file, ask it to identify the date column, explain missing values and calculate totals before drawing a trend line. Independently check a few rows. A polished chart can still be based on the wrong currency, duplicated records or an incomplete month.
Free access, paid plans and GPTs
Free accounts have access to tools such as search, uploads, image creation, data analysis and existing GPTs, with applicable limits. Tool quotas can differ from ordinary chat limits, and the interface reports when a limit is reached. A paid plan is not a promise that every tool is unrestricted.
OpenAI’s November 2023 GPTs announcement introduced creating tailored versions of ChatGPT through instructions, knowledge and selected capabilities, with access initially offered to Plus and Enterprise users. Using a GPT and building one are distinct activities. Check the account’s available tools before choosing a subscription; a plan name alone is not a promise of unlimited use.
Privacy controls are separate from paying
Do not assume a personal paid subscription automatically excludes conversations from model improvement. In Settings, Data Controls, the “Improve the model for everyone” preference controls that use. Turning it off leaves ordinary conversations in history while excluding them from training; the preference is available to signed-in users and also through controls for signed-out use. Team and Enterprise workspace rules can differ. See the Data Controls FAQ.
Temporary Chat is a separate choice. OpenAI’s February 2024 announcement describes chats that do not appear in history, do not use memory and are not used for model training. Memory, chat history and training preferences are different controls; deleting a chat does not by itself erase a saved memory. Information sent to an external service also falls under that service’s own policies.
Build a small, repeatable routine
Pick one useful task, give the necessary context, state the output you need and decide how you will verify it. Keep the successful brief as a starting point, then adjust it when the source material changes. That routine is more durable than memorising a model menu or a quota that may change next month.
A feature menu will keep changing. For this April 2025 guide, the durable lesson is to understand which tool is handling the task, what information you supply and how you will check the result. A useful workflow should survive a rearranged menu.
