AI Tech Deck

Best AI chatbot in 2026: ChatGPT vs Claude vs Gemini vs Perplexity

Four assistants now cost the same $20 a month at the tier most people buy, and the free tiers have never been more generous. The differences are in what the $20 actually buys, how much text each one can hold at once, and what happens to your chats. Here is the table, then one pick for each kind of user.

By HarryChecked September 2026

Four glossy playing cards fanned on a dark navy table, each card showing a different glossy speech-bubble shape, lit from one side
AI-generated illustration

Plans, prices and models were checked against each vendor's pricing and help pages in September 2026. Some vendor links on this site are affiliate links; they never affect a pick. See our disclosure and method.

The comparison table

As of September 2026ChatGPTClaudeGeminiPerplexity
Current modelsGPT-6 Astra, GPT-5.6 Sol, Terra and LunaClaude Fable 5.1, Opus 5, Sonnet 5, Haiku 4.5Gemini 3.1 Pro, Gemini 3.8 Flash and 3.6 Flash, Deep ThinkSonar 2 plus third-party models: GPT-5.6 Terra, Claude Sonnet 5, Gemini 3.7 Flash, Grok 4.6 and others
Free tierUnlimited text chat on GPT-5.6 Luna; limited uploads, images, voice, deep research and CodexSonnet 5 and Haiku 4.5, web search, files and code, memory, voice, connectors, up to 5 projectsGemini 3.6 Flash with varying access to 3.1 Pro, image generation, Deep Research, Gemini Live, 15 GB storagePractically unlimited basic searches, 3 Pro Searches a day, 1 Research query a month, no advanced models
Cheapest paidGo, about $8/mo; may show adsPro, $20/mo ($17/mo billed annually)AI Plus, $4.99/mo, 2x free limits, 400 GBEducation Pro, $10/mo with student verification; otherwise Pro at $20
What $20 gets youPlus: GPT-6 Astra and GPT-5.6 Sol, projects, tasks, custom GPTs, expanded Codex, ChatGPT WorkPro: Opus 5 plus Fable usage credits, Claude Code, Research, Design, Slides and Docs, Claude in Chrome and Microsoft 365AI Pro ($19.99): 4x limits, expanded 3.1 Pro and Deep Research, Gemini Spark, Chrome auto browse, 5 TB, YouTube Premium LitePro: model switching, near-unlimited Pro Search, image generation, file uploads, Perplexity Computer, Comet
Context (what it can hold)Instant 27K free, 54K Plus, 128K Pro; reasoning 256K Plus, 400K ProUp to 1M tokens, varies by model, on all plans32K free, 128K AI Plus, 1M on AI Pro and UltraDepends on the model chosen; Perplexity does not publish a plan-level figure
VoiceYes on every plan; voice with video from Go upYes on every plan including FreeGemini Live on every planVoice in the apps and a Voice Mode in Perplexity Computer
IntegrationsApps for Excel, Word, PowerPoint and Sheets, connectors, built-in browser, Codex, ChatGPT WorkConnectors and MCP, desktop extensions, Claude in Chrome, Microsoft 365 and Outlook, Claude CodeGmail, Docs, Sheets, Chrome, Photos, Android, Antigravity and Jules for codersComet browser, Perplexity Computer, email, file connectors
Trains on your chats by default?Yes on Free, Plus and Pro; opt out under Data ControlsYes unless you opt out under Privacy; Team and Enterprise neverYes, with human review of samples; controlled by Keep ActivityYes, AI data retention is on by default; opt out in Preferences

Prices are US list prices. Google's page shows $19.99 for AI Pro and $99.99 or $199.99 for the two Ultra tiers; ChatGPT Pro is $100 or $200; Claude Max starts at $100; Perplexity Max is $200. All four charge tax on top where it applies.

The one thing that changed this year: every vendor now meters usage rather than counting messages. Gemini moved to compute-based limits in May, Claude resets a rolling five-hour window with weekly caps on top, and ChatGPT's tiers are described as 5x and 20x rather than as message counts. The practical effect is that a long document or a reasoning model burns your allowance faster than a short chat, on every service.

Writers: Claude

Claude is the pick if the output is prose someone else will read. Sonnet 5 is on the free tier and produces the cleanest short-form drafts of the four with the least editing, and Pro at $20 adds Opus 5, which is the model to use for a long document that has to hold its structure. Every plan gets up to a million tokens of context depending on model, so you can paste a whole manuscript and ask about chapter three without the assistant losing chapter one. Claude Docs, Design and Slides on Pro cover the formatting step. The catch is volume: usage resets on a five-hour window and heavy users hit it, which is why the $100 Max tier exists. If you write all day, price Max against ChatGPT Pro before choosing.

Students: ChatGPT, with two exceptions

ChatGPT's free tier is the deepest: unlimited text on GPT-5.6 Luna, study mode, projects and limited Codex, all at $0. For most coursework that is enough, and the roughly $8 Go plan lifts the upload and voice limits when you hit them. Two exceptions. Perplexity's Education Pro is $10 a month with student verification and gives you the full Pro plan, which is the better deal if your work is research-heavy. And Google's student offer, one year of AI Pro free with 5 TB of storage and a 1M-token context window, is hard to beat if you qualify; check the eligibility page rather than assuming.

Google households: Gemini

If your email is Gmail, your files are in Drive and your phone is Android, Gemini is the assistant that is already where your work is. The free tier is real: Gemini 3.6 Flash, varying access to 3.1 Pro, Deep Research and Gemini Live. AI Pro at $19.99 is the strongest $20 bundle in the category on paper: 4x the free limits, the full 1M context, Gemini in Gmail, Docs and Sheets, Chrome auto browse, Gemini Spark for hand-off tasks, 5 TB of storage and YouTube Premium Lite. AI Plus at $4.99 with 400 GB is the cheapest paid step of any of the four. The trade-off is the model: in our sessions Gemini's long-form writing still needs more editing than Claude's or ChatGPT's, and the privacy notice says human reviewers read samples of chats unless you turn Keep Activity off.

Researchers: Perplexity

Perplexity is built around showing its sources, and for fact-finding it is still the fastest of the four to a cited answer. Pro at $20 lets you switch the model behind the answer, including GPT-5.6 Terra, Claude Sonnet 5, Gemini 3.7 Flash and Grok 4.6, and it includes the Comet browser and Perplexity Computer, an agent that turns a request into a finished file. Max at $200 adds GPT-5.6 Sol and Claude Opus 5 and much higher Computer limits. The free tier's three Pro Searches a day and one Research query a month are the tightest caps in this table, so a serious researcher pays. Note the Enterprise plans are the only Perplexity tiers where data is never used for training; on Pro you opt out in settings.

Developers: ChatGPT or Claude, decided by your editor

Both include a coding agent in the $20 plan: ChatGPT Plus expands Codex and ChatGPT Work, and Claude Pro includes Claude Code. ChatGPT's argument is GPT-6 Astra on Plus, the strongest reasoning model any of the four offers at $20. Claude's argument is the 1M-token context on every plan and Claude Code's terminal-first workflow. If you already use Cursor, pick the assistant whose models you like there, since you will run the same models in both places. Gemini's AI Pro is the value pick for coders on Google's stack, with expanded limits in Antigravity, Jules and Android Studio plus $10 a month of Cloud credits.

Budget: free, then Google's $4.99 or ChatGPT's $8

Stay free as long as you can. ChatGPT free covers unlimited everyday chat, Claude free covers writing on Sonnet 5 with voice and memory, and Gemini free covers Google-app tasks. When you pay, the two cheap steps are Google AI Plus at $4.99 (2x limits, 400 GB) and ChatGPT Go at about $8 (higher limits, longer memory, may show ads). Perplexity has no cheap tier without a student ID, and Claude's first paid step is $20. Do not buy a $20 plan to fix a limit you hit twice a month; buy it the week you notice you are waiting for a reset to finish real work.

Verdict: ChatGPT is the default, and three specific people should pick something else.

Writers pick Claude. Google Workspace households pick Gemini. People whose questions are mostly research questions pick Perplexity. Everyone else starts on ChatGPT free, moves to Plus when they hit limits on real work, and turns off the training toggle on day one whichever card they pick.

How we compared them

We read each vendor's pricing page and help center in September 2026 and pulled the plan names, prices, models and limits from those pages; where a vendor's page hides the price behind a region selector we used the US figure reported by at least two independent guides checked the same month. We then ran the same five tasks in each free tier and each $20 tier: a 900-word draft, a long PDF summary, a cited research question, a code fix and a voice conversation. Our method page explains the weighting. The model tracker has the API prices and context windows behind each assistant, and free tiers and deals tracks the student and promotional offers.

Quick answers

Which AI chatbot is best overall in 2026?

ChatGPT, because its free tier gives unlimited text chat on GPT-5.6 Luna and its $20 Plus plan adds GPT-6 Astra. Claude is the better pick if you write for a living, Gemini if you live in Google Workspace, and Perplexity if most of your questions are research questions.

Which AI chatbot has the best free tier?

ChatGPT for volume, since text chat on GPT-5.6 Luna is unlimited. Claude's free tier is smaller but includes Sonnet 5, voice mode, memory and connectors. Gemini's free tier bundles 15 GB of storage and varying access to Gemini 3.1 Pro. Perplexity's free tier is strong for search but caps Pro Search at three a day.

Do these chatbots train on my conversations?

On personal plans, yes by default for all four as of September 2026, and each offers an opt-out in settings: ChatGPT under Data Controls, Claude under Privacy, Gemini under Gemini Apps Activity, Perplexity under AI data retention. Business and API plans do not train by default.

Sources

More guides