ChatGPT

Learning Objectives

• Understand ChatGPT's core capabilities and how they differ across pricing tiers
• Identify the specific task types where ChatGPT outperforms alternatives
• Apply practical strategies for getting the best results from ChatGPT

What Is ChatGPT?

ChatGPT is OpenAI's consumer AI chat interface — the product that launched the modern AI era when it reached one million users in five days after its November 2022 release. At roughly one billion weekly active users, it remains the most widely used AI product on the planet.

Under the hood, ChatGPT runs GPT-6 Astra on paid tiers, released September 3, 2026. Below it sits the GPT-5.6 family, which splits into three named capability tiers rather than a single model: Sol the flagship, Terra the balanced middle, and Luna the fastest and cheapest. Plus, Pro, Business, and Enterprise users get Astra; since August 2026, Free and Go users get GPT-5.6 Luna as their default along with unlimited text conversations. OpenAI's valuation reached $730 billion in early 2026, with $168 billion in total funding, as the company restructured as a Public Benefit Corporation (PBC).

Pricing Tiers

The free tier changed meaningfully in August 2026: text conversations are now unlimited, so the reason to pay is no longer message volume. Go at $8/month raises the remaining caps on files, images and voice. For professionals, Plus at $20/month is the entry point to the GPT-6 Astra flagship, GPT Image 2, and Deep Research. Business ($25 to 30/user/month) adds team workspace and admin controls. Since August 10, 2026 Business also offers premium seats, which carry five times the usage of a standard seat and remove the five-hour usage limit; seats are assignable per teammate, so a team can buy heavy capacity only for the people who need it rather than upgrading everyone.

What pays for that free tier is now a real business rather than an experiment. On August 31, 2026 OpenAI reported that ChatGPT Ads had reached an annualized run rate of $1 billion in fewer than 200 days, across tens of thousands of advertisers in more than 40 countries, and it describes advertising as one pillar of a diversified model alongside subscriptions, enterprise deals and usage-based API billing. That matters when you are choosing a tier: the unlimited free plan exists because it is monetized, not in spite of it.

Sponsored Agents — an ad you can interrogate

On September 16, 2026 OpenAI began testing Sponsored Agents: after clicking an ad, a user can open a separate, clearly labeled conversation with an agent the advertiser pays for, ask follow-up questions about fit, sizing or care, and then follow a link to the business. OpenAI's stated boundary is that this conversation is distinct both from ChatGPT's own answers and from the chat the user started in — the ad does not enter the thread you were already having. Testing is limited to selected advertisers in the United States, so treat it as a pilot rather than a shipped tier feature. The same announcement let advertisers build campaigns from a prompt in ChatGPT Work, added AI creative tools to Ads Manager, and named HubSpot and Shopify as the first partners in customer relationship management and ecommerce, which moves ChatGPT Ads into the stack businesses already run rather than requiring a separate destination.

The direction matters more than the pilot. An ad that answers questions is a different object from a sponsored link, and it puts the labeling promise under more strain than static placements did: the boundary between a paid agent and the assistant's own judgment is one the user has to be able to see, every time, in a medium where both are conversation. Whether that holds is the thing worth watching.

Core Features

GPT-6 Astra — The Underlying Model

ChatGPT's paid tiers run GPT-6 Astra, released September 3, 2026 — OpenAI's first new GPT generation in more than a year. Pro, Business and Enterprise also get GPT-6 Astra Pro. Astra's strongest reported gains are in computer use and long agentic runs, and it is the first model OpenAI has rated Critical for cybersecurity, so it refuses advanced offensive-security work.

Below it, the GPT-5.6 family remains in service. It changed the naming convention: the number marks the generation, while Sol, Terra, and Luna name durable capability tiers that each advance on their own cadence, all carrying a roughly 1.05 million token context window. Because Astra costs double Sol's API input rate, GPT-5.6 is still where OpenAI's cost-efficient options live.

On August 6, 2026 OpenAI moved Free and Go users onto GPT-5.6 Luna — the fastest and most cost-efficient tier — and made text conversations unlimited on those tiers. It also added a Think button that spends more reasoning on a hard question on demand. OpenAI reports Luna makes 62 percent fewer factual errors than the GPT-5.5 Instant model it replaced, measured on its own internal evaluations. Paying subscribers received an updated Sol at the same time, and moved to Astra a month later.

ChatGPT for Teens

Beginning August 18, 2026, and rolling out globally over about two weeks, OpenAI began routing under-18 users into ChatGPT for Teens, a restricted mode rather than a separate product or paid tier. Teenagers do not opt in: anyone who states an age of 13 to 17, or whom OpenAI's age-prediction system estimates is under 18, is moved across automatically, with no new account and nothing to switch on.

In that mode the model will not use romantic language, imply it has feelings or consciousness, call itself a friend, or encourage emotional dependence, and it shows more frequent reminders that it is artificial intelligence. A Study Hours setting activates Study Mode on a schedule, and the system redirects a teenager who appears to be trying to skip an assignment. Parents get controls over settings, safety notifications, and Quiet Hours. Learning visualizations now cover more than 300 topics, up from 70 at launch.

GPT Image 2 — Image Generation

Image generation is built directly into conversation — no separate tool needed. Describe an image in plain language, refine it through follow-up prompts, and iterate. GPT Image 2 (April 21, 2026) is OpenAI's first reasoning-native image model, with character-level multilingual text rendering (Japanese, Korean, Chinese, Hindi, Bengali), web-search grounding for real-time facts, and 2K resolution at up to 8 images per prompt.

Image Provenance — C2PA Metadata and Google's SynthID Watermark

OpenAI shipped two complementary provenance signals on every AI-generated image from GPT Image 2 and ChatGPT's image surface in May 2026. C2PA metadata — the open Coalition for Content Provenance and Authenticity standard — embeds a cryptographically signed manifest in each image identifying it as AI-generated, including the model that produced it. SynthID — Google's invisible watermark, adopted by OpenAI through a cross-lab agreement — encodes the same provenance signal directly into the image pixels in a way designed to survive screenshots, cropping, resizing, and common social-media transformations.

OpenAI also previewed a public verification tool that detects both C2PA metadata and SynthID watermarks. The tool initially recognizes only OpenAI-generated images; broader coverage is planned as more vendors adopt the same signals.

Voice Mode — Now Powered by GPT-Live

ChatGPT's voice experience runs on GPT-Live as of July 8, 2026, replacing the earlier Advanced Voice Mode. GPT-Live uses a full-duplex architecture — it listens and speaks at the same time, so it handles natural turn-taking, can interject short acknowledgments like "mhmm," or stay quiet while you think. Two models power it: GPT-Live-1 mini is the new default (including on the Free tier), and paid tiers get the larger GPT-Live-1. For anything that needs web search, deeper reasoning, or a complex task, GPT-Live hands off to a frontier model behind the scenes and folds the result back into the conversation. The net effect is a voice partner that feels closer to a phone call than a walkie-talkie — useful for hands-free access, language practice, or thinking through problems out loud.

Deep Research

An autonomous research mode that spends 5 to 30 minutes searching dozens of web sources, synthesizing information, and producing a comprehensive structured report with citations. For research tasks that would otherwise take hours, Deep Research is one of the most practically valuable features in any AI product.

Custom GPTs

Create specialized AI assistants with custom instructions, uploaded knowledge files, and configured tool access. The GPT Store hosts thousands of community-built GPTs for specific use cases — from legal document review to language tutoring to domain-specific coding help.

ChatGPT Health — Personal Health Assistant

On July 23, 2026, OpenAI made ChatGPT Health available to every US adult across all tiers — Free, Go, Plus, and Pro — after a six-month pilot. The feature lets you securely connect Apple Health data and pull in medical records from hospital systems like Epic and Oracle Health, plus platforms like One Medical and Function Health, so ChatGPT can answer questions about test results, medications, diet, and workouts against your actual history rather than in the abstract. OpenAI says more than 300 million people now ask ChatGPT health questions each week, up from roughly 230 million in January, and that connected health data is not used to train its models or to target ads.

ChatGPT for Healthcare — the clinician-side product

Do not confuse this with ChatGPT Health above. ChatGPT Health is the consumer feature, where you connect your own records. ChatGPT for Healthcare is a separate product sold to healthcare organizations, and on September 1, 2026 OpenAI opened an Epic integration for it. That integration pulls authorized patient context — visit notes, laboratory results, medications and specialist documentation — into ChatGPT so a clinician can ask what changed since a patient's last visit rather than reading the chart end to end. It supports two shapes: bringing EHR context into ChatGPT, and embedding ChatGPT inside an EHR layout so the work happens without leaving the patient chart.

A companion Healthcare Public Data plugin adds structured access to nine official public sources, including PubMed, ClinicalTrials.gov, DailyMed, RxNorm and Medicare coverage policy — so a team can compare trial eligibility criteria or confirm a medication's current label without searching each source separately. The University of California, San Francisco health system is a pilot partner. OpenAI says it works with physicians across 60 countries, 49 languages and 26 medical specialties, who have reviewed more than 700,000 model responses.

Canvas

A collaborative document editing workspace within ChatGPT — write documents, get AI suggestions inline, and iterate in a shared editing environment rather than through the chat message thread.

Memory — Automatic Background Synthesis

ChatGPT keeps a long-term memory that carries context across separate conversations, so you don't have to re-introduce your projects, preferences, or writing style each session. The current architecture, which OpenAI calls dreaming, runs as a background process that synthesizes what the model knows about you across many chats — without you having to explicitly tell it to remember anything. It also keeps that memory current on its own, revising a note like "planning a trip to Singapore in July" to "visited Singapore in July" once the date passes.

A roughly five-times reduction in the cost of running this synthesis let OpenAI extend automatic memory to the free tier, not just Plus and Pro. Everything it stores is reviewable and editable on a memory summary page, where you can add facts, delete them, or steer which topics ChatGPT should and shouldn't bring up.

Codex, ChatGPT, and the Unified Product Surface

OpenAI's product surface is converging. Co-founder Greg Brockman permanently leads product strategy, with Thibault Sottiaux — the engineer who scaled Codex into one of OpenAI's fastest-growing products — running a combined core product team across ChatGPT, the Codex coding agent, and the developer API. The stated goal is to fold all three into a single unified platform aimed at agentic workloads, with Codex's productivity capabilities reaching beyond pure coding and eventually merging with ChatGPT itself. The reorganization is widely read against a prospective OpenAI initial public offering, one Sam Altman has since pushed past 2026, and follows earlier internal decisions to discontinue peripheral projects and concentrate effort on the agentic core.

The most visible consumer-facing piece of the consolidation is Codex in the ChatGPT mobile app. Codex now runs inside the iOS and Android app in preview on every subscription tier including Free. Mobile Codex is scoped to remote review and approval (read outputs, approve commands, change models, switch threads) rather than autonomous background execution, which remains desktop-only. The feature brings Codex's sandboxed coding workflow to a phone, lowering the bar for adoption beyond the existing Codex desktop base and signaling the broader direction: one platform, agentic-first, accessible from anywhere.

The next phase of that consolidation is a planned "superapp" redesign of ChatGPT — OpenAI's biggest interface overhaul yet — that reshapes the app to steer users toward Codex coding tools, image generation, and partner services such as Canva and Booking.com. The redesign is widely read as a push toward higher-margin enterprise and developer revenue ahead of the IPO, though as of mid-2026 it is announced and rolling out rather than broadly shipped. Business customers already account for roughly 40 percent of OpenAI's revenue, a share the company expects to approach half by year-end.

Personal Finance — Bank-Account Integration via Plaid

ChatGPT Pro subscribers can connect more than 12,000 financial institutions — including Schwab, Fidelity, Chase, Robinhood, American Express, and Capital One — through Plaid and ask the model to analyze spending, portfolio performance, and financial planning across linked accounts. The integration surfaces in the sidebar or via the @Finances, connect my accounts slash command, and is restricted to the Pro tier on launch.

OpenAI's stated controls: synced transaction data is removed from ChatGPT within 30 days of disconnect, and financial memories can be viewed or deleted separately from the rest of ChatGPT's memory. The launch pulls ChatGPT directly into the territory occupied by Intuit's Mint, YNAB, and bank-native assistants — with the trade-off that a general-purpose model now reads every transaction in a user's life. For Pro subscribers willing to grant that access, it collapses spending analysis, portfolio review, and tax-style planning into the same chat thread already handling code and research.

Lockdown Mode — Prompt-Injection Defense

As ChatGPT gained web browsing, connectors, and agentic execution, it also inherited their core security risk: prompt injection, where hidden instructions buried in a webpage, document, or connected data source trick the model into acting against the user. OpenAI's answer is Lockdown Mode, an opt-in setting that walls ChatGPT off from the external data paths most likely to carry an injection. Turning it on disables live web access, Deep Research, Agent Mode, image responses, live connectors, and file downloads — trading capability for a much smaller attack surface.

Lockdown Mode is aimed at people and organizations that handle sensitive data and want stricter protection from data-exfiltration risk, and it rolls out to personal accounts and self-serve Business accounts alongside new "Elevated Risk" labels that flag higher-risk actions. OpenAI is explicit that it is a mitigation rather than a cure — an injection hidden in cached content or an uploaded file can still affect a response even with Lockdown Mode on. For anyone connecting bank accounts or sensitive corporate data through ChatGPT's agentic features, it is the most direct lever for reducing exposure.

Apple Messages Plug-In — Reading and Sending Your Texts

On August 20, 2026 OpenAI shipped an Apple Messages plug-in for the ChatGPT desktop app. It can search your message history, summarize threads, draft replies, and send iMessage, SMS and RCS texts as you. Availability is narrow: it runs only on Apple Silicon Macs and only inside the Codex and ChatGPT Work surfaces — not the web app, not mobile, not the command-line tool or the editor extension. Installing it means granting ChatGPT broad macOS permissions for automation, contacts and full disk access.

OpenAI for Countries — Malta National Rollout

Malta has become the first country in the world to make ChatGPT Plus a citizen benefit, under OpenAI's new OpenAI for Countries program. Every Maltese citizen and registered resident can claim a free one-year ChatGPT Plus subscription after completing the AI for All online course developed by the University of Malta, with the Malta Digital Innovation Authority managing distribution. The program is OpenAI's inaugural national-scale deal under the OpenAI for Countries initiative — a distribution model that anchors sovereign AI adoption around ChatGPT and reframes consumer AI access as a public-good utility rather than a per-user subscription.

For learners outside Malta, the deal is a useful signpost: governments increasingly view consumer-grade AI access as infrastructure on par with libraries or public broadband. Expect more national-scale programs to follow through 2026, and expect their terms — tier coverage, literacy requirements, data residency — to vary widely from country to country.

ChatGPT EDU — Higher-Education Enterprise Tier

ChatGPT EDU is OpenAI's higher-education-specific deployment tier, sold to universities at organization scale rather than per-seat consumer pricing. The marquee deployment is California State University, which renewed its systemwide OpenAI partnership in May 2026 at $13 million a year for three years — extending ChatGPT EDU access to roughly 470,000 students and 63,000 faculty across 23 campuses. It is the largest deployment of ChatGPT inside any single organization to date.

A 2025 systemwide CSU survey found 95% of students using an AI tool and 84% using ChatGPT specifically, but 82% of students also said they worry AI will hurt their future job security. A faculty petition opposing the contract continues, and campuses adopted AI tools in the first year without consistent guidance or training, surfacing recurring concerns about data privacy, academic integrity, and equitable access.

Trusted Contact Safety Safeguard

OpenAI ships an opt-in Trusted Contact feature for ChatGPT that lets adult users designate a friend or family member to receive an alert if the system detects suicidal ideation in a conversation. The launch followed Pennsylvania's lawsuit against Character.AI over a chatbot posing as a licensed psychiatrist and broader industry pressure on chatbot self-harm handling.

How it works:

• An adult user opts in under Settings and names one or more contacts (email, phone, or another ChatGPT account).
• When OpenAI's safety system flags a message as potential suicidal ideation, the conversation is escalated to OpenAI's human safety review team, which aims to act in under one hour.
• If reviewers judge a serious safety risk, the designated contact receives a check-in nudge by email, text, or push notification.
• The alert does not include the chat content itself — only that someone the contact knows may be at risk and would benefit from a wellness check.

The Trusted Contact launch sits inside a broader 2026 pattern of frontier AI labs adding self-harm-specific guardrails — Anthropic has published sycophancy-reduction research, and Pennsylvania's Character.AI suit has pushed the entire chatbot industry to harden mental-health-adjacent conversation handling.

The Development Company — Enterprise Services Joint Venture

OpenAI runs "The Development Company" — an enterprise AI services joint venture raising $4 billion at a $10 billion valuation from TPG, Brookfield Asset Management, Advent, and Bain Capital, alongside 15 other investors. Modeled on Palantir's forward-deployed engineer approach, The Development Company embeds OpenAI engineers into mid-sized customer organizations to integrate ChatGPT, GPT-5.6, and the Codex platform into core operations rather than selling traditional SaaS access alone. Anthropic runs a parallel joint venture with Blackstone, Hellman & Friedman, and Goldman Sachs — together signaling that direct integration services may be how mid-market frontier-AI distribution is increasingly sold.

Toward IPO — Confidential Filing and NVIDIA's $30 Billion Commitment

OpenAI is preparing to file confidentially with the SEC for an initial public offering, targeting a September 2026 listing with Goldman Sachs and Morgan Stanley leading the underwriting. The filing arrives alongside a parallel SpaceX initial public offering, setting up a head-to-head financial event between two of the year's most-watched offerings. Valuation and float details have not yet been disclosed.

Among the structural facts now public ahead of the offering: NVIDIA holds a $30 billion commitment to OpenAI — the largest single investment in private-AI history — paired with capacity expansion that NVIDIA CEO Jensen Huang has described as "quite significant" on NVIDIA's Q1 fiscal 2027 earnings call. The commitment sits alongside Microsoft's $13 billion-plus investment and Apple's iOS integration, anchoring a multi-layer compute and distribution stack heading into the public-market debut. The recently-resolved Elon Musk lawsuit against OpenAI — which a California jury rejected unanimously on statute-of-limitations grounds in May 2026 — removed a multi-year governance overhang ahead of the filing.

Strengths

• Broadest ecosystem: More integrations, tutorials, plugins, and community resources than any competitor — most useful when you're learning or troubleshooting
• Image generation built-in: Generate and refine images in the same conversation, with one of the strongest text-in-image capabilities available
• GPT-Live voice: Full-duplex GPT-Live models (July 2026) that listen and speak at once for natural, phone-call-like conversation
• Deep Research: Highly capable autonomous research agent for comprehensive topic reports
• Custom GPTs: Customize your experience or leverage thousands of community-built specialized assistants
• Feature velocity: OpenAI releases new capabilities frequently; Plus and Pro subscribers get early access

Limitations & Considerations

• Context window in chat: While the API supports 1 million tokens, the chat interface has a shorter effective context — for very long document analysis, Claude (1 million) or Gemini (1 million chat) may be more practical
• Coding benchmarks: Claude Fable 5 leads SWE-bench Pro at 80.3% for the hardest end-to-end software-engineering tasks, and OpenAI has not published a SWE-bench Pro score for GPT-6 Astra or GPT-5.6 Sol — a conspicuous gap given that GPT-5.5 trailed at 58.6%. OpenAI Codex remains its dedicated coding product
• Pro tier cost: At $200/month, the Pro tier is only worthwhile for very heavy professional use of extended reasoning and unlimited Deep Research
• Data privacy (Free/Plus): Conversations may be used to improve OpenAI's models by default — opt out under Settings → Data Controls, or use Team/Enterprise for contractual data protection
• EU regulatory status: On August 31, 2026 the European Commission designated ChatGPT a very large online search engine under the Digital Services Act, the first time a general-purpose AI assistant has been brought into that regime in its own right. OpenAI has four months, until January 2027, to meet obligations including systemic-risk assessment, independent annual audits and researcher access to data. This is a separate law from the EU AI Act, which governs model providers; expect visible changes to how ChatGPT handles illegal content, minors and ad targeting for European users

Best Use Cases

When to choose alternatives:

• Very long document analysis → Claude (consistent 1 million context in chat)
• Complex software engineering → Claude Fable 5 or OpenAI Codex
• Source-cited research → Perplexity (every claim linked to source)
• Google Workspace workflows → Gemini (native integration)

Getting Started

1. Go to chat.openai.com — create a free account with email or Google sign-in
2. Start on the free tier: GPT-5.6 Luna with unlimited text chats is genuinely capable for most tasks
3. Try image generation — describe a scene and iterate through follow-up prompts
4. Explore the GPT Store (top-left menu) to find specialized assistants for your use cases
5. Upgrade to Plus if you want Deep Research mode, higher limits, and full Voice Mode
6. Adjust data privacy under Settings → Data Controls to opt out of model training if preferred

Key Takeaways

• ChatGPT is the world's most-used AI product with the broadest ecosystem of integrations, tutorials, and community resources — the natural starting point for AI beginners
• ChatGPT runs GPT-6 Astra on paid tiers since September 3, 2026, with GPT-5.6 Luna still the Free and Go default — unlimited text conversations on those tiers and a Think button for harder questions
• GPT Image 2 makes ChatGPT the strongest all-in-one choice when image generation is part of your workflow — the first reasoning-native image model with built-in multilingual text rendering and web grounding
• Deep Research is one of the most practically useful features in any AI product — capable of producing comprehensive reports that would otherwise take hours
• ChatGPT for Healthcare is the clinician-side product and is distinct from consumer ChatGPT Health: since September 1, 2026 it connects to Epic environments to pull authorized patient context into ChatGPT (or embed ChatGPT in the chart), with a plugin covering nine official public sources; UCSF Health is a pilot partner
• ChatGPT Health opened to every US adult across all tiers on July 23, 2026 — connect Apple Health and hospital records (Epic, Oracle Health, One Medical, Function Health) for guidance grounded in your own history; 300 million people now ask health questions weekly, but OpenAI is clear it is not a substitute for a clinician
• ChatGPT voice now runs on GPT-Live (July 8, 2026) — full-duplex GPT-Live-1 and GPT-Live-1 mini models that listen and speak simultaneously, handle turn-taking, and hand off to a frontier model for harder questions, replacing Advanced Voice Mode
• The opt-in Trusted Contact safeguard (May 7, 2026) lets adult users name a friend or family member to receive a check-in nudge when OpenAI's safety team flags suicidal ideation in a conversation — under one-hour human review, without sharing chat content
• Image provenance ships on every GPT Image 2 generation as of May 2026: C2PA metadata identifies AI-generated content, and Google's SynthID watermark — adopted by OpenAI in a cross-lab agreement — encodes the same signal directly in pixels to survive screenshots and resizing; a public verification tool is in preview
• Pro subscribers can now connect bank accounts and brokerages across 12,000+ institutions via Plaid for in-chat spending analysis, portfolio review, and financial planning — with line-item transaction visibility, 30-day deletion after disconnect, and a real privacy trade-off to weigh before granting access to large balances
• Malta is the first country to offer free ChatGPT Plus to every citizen and registered resident, paired with a University of Malta AI literacy course — the inaugural deal under OpenAI's OpenAI for Countries national-distribution program
• California State University renewed its ChatGPT EDU contract at $13 million a year for three years — extending OpenAI's largest single-organization deployment across 470,000 students and 63,000 faculty across 23 campuses; a 2025 survey found 95% of students using AI but 82% worrying it will hurt their job security
• OpenAI's product surface is converging under Brockman: ChatGPT, Codex, and the developer API are being folded into one unified agentic platform, with Sottiaux running the combined product team and Codex mobile already shipping on every tier
• For complex coding or very long document analysis, specialized alternatives (Claude Fable 5, OpenAI Codex) often outperform on those specific benchmarks
• The Plus tier at $20/month is the best value entry point; Pro at $200/month is for very heavy professional use only