There is no single best assistant in 2026. ChatGPT, Claude and Gemini are all good at everyday questions, and the real differences show up in specific jobs. Pick ChatGPT if you want one app that does the most things (voice with video, image creation, scheduled tasks, plugins). Pick Claude if you work with long documents or careful writing: every plan, including Free, lists a context window of up to 1M tokens depending on the model. Pick Gemini if your work already lives in Gmail, Docs and Drive, or you want Google Search behind the answer. Pick Perplexity when you need sources first. The way you prompt them is mostly the same, with a few habits that suit each one, covered below with one prompt adapted three ways.
ChatGPT vs Claude vs Gemini at a glance
All three have a free plan, a roughly $20 plan and a $100-plus plan for heavy use. The table compares what each vendor's own pages said when I checked them on 20 October 2026. Plans and limits change often, so treat it as a snapshot.
| ChatGPT | Claude | Gemini | |
|---|---|---|---|
| Free plan model | Luna models; no GPT-6 Sol (OpenAI) | Sonnet and Haiku | "Flash-Lite" model |
| Main paid plan | Plus, $20/month | Pro, $20/month | Google AI Pro, $19.99/month |
| Context on the free plan | 27K tokens for GPT Instant, about 12 pages of input | Up to 1M tokens, depending on model | Not stated; 1M tokens on AI Pro and Ultra |
| Web search | All plans | All plans | Yes, with sources and a Google Search double-check |
| Voice | All plans; video from Go up | Voice mode (beta) on all plans | Gemini Live on the free plan, in the mobile app |
| Images | Image creation on all plans, limited on Free | Reads images; models output text only | Image generation and editing on Free; video from AI Plus |
| Projects and memory | Projects on all plans; memory limited on Free | Up to 5 projects on Free; memory on all plans | Gems on Free; memory of past chats for personal accounts |
Sources: ChatGPT pricing, What is ChatGPT Plus, Claude pricing, Gemini subscriptions, Gemini file uploads, Gemini sources.
The context row matters more than most people realise. On ChatGPT's free plan, the input maximum for GPT Instant is listed as about 12 pages of text, so a long contract or report gets cut down before the model reads it. Claude's free plan lists up to 1M tokens. Gemini's help page says you need AI Pro or Ultra for "a larger 1 million token context window."
What is each assistant best at?
ChatGPT has the broadest toolkit. Its pricing page lists voice on every plan, voice with video from Go upwards, image creation, scheduled tasks, custom GPTs and plugins. If you want one app to plan a trip, read a photo of a menu and draft an email, it is the easiest single place to do all of it. The trade-off is that the free plan's context is small, and GPT-6 Sol, which powers GPT-6's thinking levels, starts at Plus.
Claude is the one I reach for when the input is long or the instructions are fiddly. Anthropic's own docs say Claude "responds well to clear, explicit instructions" (prompting best practices), and in my use it keeps to the word limits and formats I set. Its models take images as input but output only text (models overview), so there is no image generation, and voice mode is still labelled beta.
Gemini makes most sense if you already use Google's apps. Every paid plan puts Gemini inside Gmail, AI Pro and Ultra add Docs, and the free plan includes Deep Research and Gemini Live. Answers can also show the web pages they drew on. On the free plan you get the Flash-Lite model, so the paid tiers are a bigger step up here than on the other two.
How much do ChatGPT, Claude and Gemini cost?
The entry paid plans cost about the same in the US. As of 20 October 2026, OpenAI's help centre lists ChatGPT Plus at $20 a month and Pro tiers at $100, $200 and $500 (Pro plans). Claude lists Pro at $20 a month, or $17 a month if you pay $200 for the year, and Max from $100 (Claude pricing). Google lists AI Plus at $4.99, AI Pro at $19.99 and Ultra from $99.99 (Gemini subscriptions).
Prices are set per country. Google's page for India, for example, lists AI Plus at ₹399 and AI Pro at ₹1,950 a month. ChatGPT's cheaper Go plan sits between Free and Plus, and OpenAI's pricing page notes it "may include ads". Check the vendor's page in your own country before you pay.
Which versions do you get on each plan?
Free plans run smaller or faster models, which is worth knowing before you judge an assistant by its free tier. On ChatGPT, the free plan gets OpenAI's Luna models: the pricing page lists GPT-6 Luna, and the help centre says GPT-5.6 Luna "is becoming the default model for Free and Go users." GPT-6 Sol starts at Plus, and GPT-6 is still rolling out to paid plans. On Claude, the free plan gets Sonnet and Haiku, Pro adds Opus, and Max adds Fable (Pro can reach it through usage credits); Anthropic's model docs list the current versions as Claude Opus 5.5, Sonnet 5.5, Haiku 5.5 and Fable 5.1 (models overview). On Gemini, Free gets Flash-Lite, AI Plus gets Flash, and AI Pro and Ultra list the Pro model and Gemini 4 Argon. If an answer on a free plan feels thin, a sharper prompt often closes more of the gap than switching apps.
How is prompting different on each assistant?
The basics are identical: say what you want, who it's for, what shape the answer should take and how long it should be. All three vendors say some version of "be clear and specific." The differences are habits each one rewards.
On ChatGPT, OpenAI's guide is short: be clear and specific, refine over a few turns, and name the tone with adjectives such as "formal" or "friendly" (OpenAI prompting guide). On the free plan, keep pasted material short because of the input cap.
On Claude, put long material first and the question last. Anthropic says queries at the end "can improve response quality by up to 30 percent in tests," especially with several documents. It also suggests explaining why you want something, and wrapping each part of a complex prompt in its own tag, such as <report>.
On Gemini, Google's Workspace guide groups a good prompt into persona, task, context and format, and says the task's verb "is the most important component of a prompt." It also reports that the most fruitful prompts average around 21 words, while people usually try fewer than nine (Gemini prompting guide). Google's developer docs add that newer Gemini models default to direct answers, so ask for detail if you want it (prompt design strategies).
One prompt, adapted for ChatGPT, Claude and Gemini
Here is a request I see all the time, in its usual form:
Summarise this survey and tell me what matters.
None of the three can do much with that. They don't know who "me" is, what "matters" means or how long the answer should be. Below is the same request, written once for any assistant and then adjusted for each.
| Assistant | Prompt |
|---|---|
| Any | Summarise the attached Q3 customer survey for our product team. Give five bullets: the top complaint, the top feature request, the number that changed most since Q2, one risk, and one thing to do next sprint. Under 150 words. Only quote customers if the quote is in the file. |
| ChatGPT | The same prompt, plus "Plain, direct tone." On the free plan, attach only the results section if the full file runs past about 12 pages, and say so: "This is the results section only." |
| Claude | Paste the survey first inside <survey> tags, then the instructions, then the question last. Add the reason: "We're choosing what goes into next sprint, so rank by how many customers mention it." |
| Gemini | Start with a persona and a verb: "You're a product analyst. Summarise the attached Q3 survey for our product team." Then the five bullets and the 150-word cap. Ask for one sentence of explanation per bullet, since Google says newer Gemini models default to direct answers. |
Most of the improvement comes from the "Any" row. The per-assistant tweaks matter less, and if you only remember one rule, make it the five bullets and the word cap. What a good prompt looks like walks through more examples like this.
Where does Perplexity fit?
Perplexity is the search-first option. Every answer comes with numbered citations linking to the pages it used (how Perplexity works), so it suits questions where you need current facts and want to check the source: prices, recent research, news. Prompt it like a research question with a time frame and a source type, for example "peer-reviewed studies from the last three years." The specifics change which sources it pulls, as this guide to prompting Gemini and Perplexity shows.
Using more than one assistant without rewriting every prompt
Plenty of people use two or three of these: ChatGPT for quick questions, Claude for documents, Perplexity for research. The prompt habits above transfer between them, which means the slow part is adding the missing details each time. A pre-prompt step does that before you hit Send.
PrePrompt is a Chrome extension that adds a Rewrite button to ChatGPT, Claude, Gemini and Perplexity. It launches soon with 30 free rewrites a month, no card required; join the waitlist to hear the day it goes live. It grades your prompt first and leaves an already-clear one alone.
TL;DR
- No assistant wins everything. ChatGPT has the most features, Claude handles long input on every plan, Gemini fits Google's apps, and Perplexity is for cited search.
- Entry paid plans cost about $20 a month in the US for all three (checked 20 October 2026); Gemini also has AI Plus at $4.99.
- Free plans run smaller models, and ChatGPT's free input cap is about 12 pages.
- One clear prompt works everywhere. Small tweaks: long material first on Claude, a persona and a verb on Gemini, a tone word on ChatGPT.
About the author. Tanuj writes PrePrompt's comparison and how-to coverage and uses ChatGPT, Claude, Gemini and Perplexity every day while testing prompt rewrites. PrePrompt's rewriter runs on a Claude Haiku model, which is worth knowing when you read a comparison from us; the extension works the same way on all four sites, and this post recommends each assistant only for the jobs it fits.
Sources: ChatGPT Plans and Pricing (openai.com); GPT-6 and other models in ChatGPT, What is ChatGPT Plus, About ChatGPT Pro plans and Prompt engineering best practices for ChatGPT (OpenAI Help Center); Claude pricing (claude.com); Models overview and Prompting best practices (Claude Platform Docs); Google AI subscriptions (gemini.google, US and India pages); Upload and analyze files and View related sources (Gemini Apps Help); Gemini for Workspace Prompting Guide 101 (Google, 2024); Prompt design strategies (Google AI for Developers); How does Perplexity work (Perplexity Help Center). All vendor pages checked 20 October 2026.
Related reading: what a pre-prompt is, how to rewrite prompts on Gemini and Perplexity and what a good prompt looks like.
