ChatGPT review: the most complete AI assistant, now with ads on the cheap plans

ChatGPT is OpenAI’s general AI assistant. It runs on the web, Windows, macOS, iOS and Android, and it answers questions, writes and edits text, analyses files, searches the web and runs agent tasks. OpenAI launched it in November 2022. It’s still the assistant most people measure every rival against.

On the Cybernock research rubric ChatGPT scores 8.5/10 and ranks third in our best AI assistants list, a hair behind Claude on the unrounded score (8.495 against 8.54). It has the widest feature set of any assistant we rank and the most flexible price ladder, running from an $8 Go plan to a $500 Pro tier. Two findings cost it points. Its chat models sit below the best from Anthropic and Google on independent leaderboards. And the Free and Go plans now carry ads.

How does ChatGPT work in October 2026?

ChatGPT now has three modes. Chat is the familiar assistant. Work, launched in July 2026, is an agent that builds finished documents, spreadsheets, slide decks and websites across your files and apps (Reuters via BNN Bloomberg). Codex is OpenAI’s coding agent.

The models changed twice in two weeks. OpenAI released the GPT-6 family on September 22, 2026: Astra (the largest), Sol and Luna. On October 7 it began moving the Chat tab to GPT-6. Plus, Pro, Business and Enterprise get GPT-6 Sol, while Free and Go users get GPT-6 Luna from October 8 (9to5Mac). The newer GPT-6.1 Sol, released on September 29, is available only in Work and Codex for now (TechCrunch).

In practice, that split matters. When you want OpenAI’s strongest model on a hard task, run it in Work or Codex rather than the Chat tab.

How good are ChatGPT’s models?

Good, but not the leaders. On the LMArena text leaderboard of October 8, 2026, OpenAI’s best entry is GPT-6.1 Sol in 21st place. Anthropic’s Claude Opus 5.5 is second and Google’s Gemini 3.8 Flash is eighth. The GPT-6 Sol model in the Chat tab ranks 67th.

The Artificial Analysis Intelligence Index tells a similar story. GPT-6 Astra scores 53 and GPT-6.1 Sol 52, against 58 for Claude Opus 5.5. The free model, GPT-6 Luna, scores 38.

So where does that leave you? ChatGPT handles everyday writing, analysis and coding well. If raw model quality is your main test, though, Claude and Gemini currently rank higher.

Which ChatGPT features matter most?

• Work mode and agents: multi-step tasks that end in a finished file or site, plus the older ChatGPT agent for browsing and form-filling.
• Codex: a coding agent included in paid plans; the Pro tiers are sized mainly by Codex usage.
• Search, deep research and file analysis: cited web answers, long research reports and analysis of PDFs and spreadsheets.
• Memory and projects: ChatGPT remembers saved details and can refer to past chats; projects group files and instructions.
• Voice and the Atlas browser: spoken conversations in the apps, and an AI browser with an agent mode.

Few rivals match this list. Gemini comes closest and adds cloud storage. Claude covers research, coding and a Chrome agent, with fewer extras. ChatGPT can also generate images; this review leaves that feature out of the score.

How much does ChatGPT cost?

Prices run from $0 to $500 a month. OpenAI’s pricing page blocked our fetches, so the prices below come from dated third-party reports, checked in October 2026.

The $100 Pro tier launched in April 2026 (TechCrunch). In September OpenAI added a $500 tier and cut the allowance for new $200 subscribers (WinBuzzer). Business pricing comes from third-party guides only. When you compare Pro tiers, read the allowance as closely as the price, because it has already shrunk once.

For most people Plus is the plan to buy, since it removes ads and opens Work and Codex. Go makes sense only if $20 is too much and you accept ads. Don’t bother with Go if ads annoy you, since a $12 gap separates it from an ad-free Plus. Skip the Pro tiers until you hit Plus limits every week.

Is ChatGPT private?

Not by default. On Free, Go, Plus and Pro, OpenAI uses your chats to train its models unless you switch off Improve the model for everyone under Settings, Data controls (TechRadar). Temporary Chats are excluded from training and deleted within 30 days. Business and Enterprise data stays out of training by default.

If you’ve ever pasted client data or health details into a chat, switch that setting off today. When you need a one-off sensitive conversation, a Temporary Chat keeps it out of training.

Ads add a second layer. Since February 2026, logged-in adults on Free and Go in the US see sponsored results chosen partly from the current conversation (TechCrunch). OpenAI says advertisers never see your chats, and you can turn off ad personalisation, but contextual ads stay. Plus and higher plans are ad-free.

One legal issue has eased. A 2025 court order in the New York Times copyright case forced OpenAI to keep deleted chats; that obligation ended on September 26, 2025, though logs already preserved are still held (Gizmodo). In March 2026 a Rome court annulled Italy’s €15 million privacy fine against OpenAI.

Where does ChatGPT fall short?

Ranked by impact, from biggest to smallest, the weaknesses line up like this.

• Model quality. The Chat-tab model sits 67th on LMArena, and even OpenAI’s best entry trails Claude Opus 5.5 and Gemini 3.8 Flash.
• Ads. Free and Go pick sponsored results partly from what you’re discussing.
• Default training. Individual plans feed chats into training until you flip one switch.
• Plan churn. Names and allowances moved several times in 2026, including a September cut to the $200 Pro allowance for new subscribers. GPT-6.1 Sol still hasn’t reached the Chat tab.

How does ChatGPT compare with Gemini, Claude and Copilot?

Gemini beats ChatGPT on price and bundles: AI Pro costs $19.99 and includes 5 TB of storage. Claude has the highest-rated models and no ads, but its usage limits run out sooner at $20. Microsoft Copilot fits better if you live in Word, Excel and Outlook. For research with a source behind every claim, Perplexity is more focused.

Who should not use ChatGPT?

Four profiles are better served elsewhere.

• Writers and analysts who judge on model quality. Claude Opus 5.5 ranks second on LMArena; ChatGPT’s best entry is 21st.
• Anyone who won’t accept ads but won’t pay $20. Gemini’s AI Plus costs $4.99 a month with no ads in the assistant.
• People who already pay Google for storage. Gemini AI Pro bundles 5 TB at $19.99.
• Office teams that live in Microsoft 365. Copilot works inside Word, Excel and Outlook.

Should you use ChatGPT?

Yes, if you want one app that does nearly everything: agents, coding, voice and research, on every platform. Pay for Plus to drop the ads and unlock the stronger models. Turn off training in Data controls if your chats include anything sensitive.

If one of the profiles above describes you, follow that pointer instead. See the full AI assistant ranking or browse all AI assistant reviews.