The short version
AI assistants now handle real work, not just questions. This guide covers 12 verified capabilities across ChatGPT, Claude, and Gemini: image editing, voice, live camera input, data analysis, agents, cited research, and app connections, with limits and concrete uses for each, checked against official documentation in November 2026.
Summarize with AI
Editorial Disclosure & Affiliate Notice
This content is published for informational and educational purposes only. It is not intended as a substitute for professional, legal, financial, or medical advice. AIUnpacker is funded through clearly labeled sponsored articles and reviews, relevant paid link placements, and affiliate commissions where disclosed. Commercial partnerships help fund the free editorial library. Sponsored work is labeled, and a payment does not change the verdict.
- For educational purposes only. Nothing here should be taken as a guarantee, recommendation, or professional recommendation.
- AI-assisted editing. Drafts are produced with AI assistance and get a human pass before they go up.
- Opinions are our own. Also, we are not affiliated with most tools we cover unless explicitly stated.
- Information may be outdated. Verify pricing, features, and policies directly with the vendor.
- Last reviewed: . Published .
Read more on our About page, Terms and Editorial Policy.
AI assistants stopped being fancy question boxes. In November 2026, ChatGPT, Claude, and Gemini can generate and edit images, hold spoken conversations, watch your camera feed, turn spreadsheets into charts, execute multi-step tasks as agents, produce cited research reports, build small apps, connect to Google Drive and GitHub, remember your preferences, and search the web with links you can check.
Video generation is not in this list. OpenAI shut down the Sora app in April 2026.
1. Generate and edit images from a sentence
Yes, assistants make pictures now, and they edit your existing photos too. ChatGPT Images works on every tier, including free, on web, iOS, and Android. You can type a prompt, pick any aspect ratio, or upload a photo and describe the change: you can even highlight an area and ask for the fix, though OpenAI warns that “highlights are not always precise, and edits may extend beyond the area you selected.”
Recent updates: Images 2.5 arrived September 8, 2026 with templates and sketch-to-image; version 2.6 landed November 4, 2026 with faster generation, saved reference photos reusable by typing @ and a name, and GIF creation.
Concrete use: upload your product photo and ask for a transparent-background version with a text banner for your shop listing. Two minutes, no design tool.
2. Have an actual spoken conversation
You can talk to these assistants out loud, interrupt them, and talk over them like a person. ChatGPT Voice runs inside the chat itself, so you can listen while reading the text and switch to typing mid-conversation. Live mode can “listen and speak at the same time,” per OpenAI’s help center, and can pull in web search, memory, and connected apps while you talk. It works on web, iOS, Android, desktop, and CarPlay.
Limits: Go and Plus plans get about 3 hours of voice in a rolling 24-hour window, Pro at $100/month gets 15 hours, and the $200 Pro tier is effectively unlimited.
Concrete use: rehearse a tough client call in the car: have it play the difficult customer, then ask what you said that sounded defensive.
3. Show your camera or screen and get live help
Gemini can watch what you’re looking at through your phone’s camera or screen and help in real time. In the Gemini mobile app, start a Live conversation, then tap the camera icon or share your screen: long-press the power button on Android to jump straight in. Google’s guidance notes Gemini doesn’t stream continuously; it processes frames periodically, so don’t expect frame-by-frame commentary.
ChatGPT has a version too: video and screen sharing remain available to eligible subscribers in the iOS and Android apps within Advanced voice mode. Claude doesn’t offer a camera feed in its consumer app.
Concrete use: point your camera at a leaking pipe and describe the problem out loud: it’s good at identifying the part and likely fix before you call a plumber.
4. Turn messy spreadsheets into charts and answers
Upload a spreadsheet and the assistant will analyze it, run statistics, and build charts: no formulas required. ChatGPT’s data analysis feature writes and runs Python in a Jupyter notebook, then returns summaries, tables, and charts. Interactive charts cover bar, line, pie, and scatter types; others come back as static images. It handles.xls,.xlsx,.csv, PDFs, and data files like.json and.xml.
Limits: the Python environment “cannot make external web requests or API calls,” so everything it analyzes must be uploaded or connected first. OpenAI also cautions that scanned PDFs may not yield reliable exact values: review the code before trusting the numbers.
Concrete use: upload a year of sales data and ask which product line’s margin is shrinking fastest: chart and reasoning in one reply.
5. Work through long PDFs and documents
Assistants read documents you upload, and the file limits are generous. Per OpenAI’s help center, documents and presentations can be up to 512 MB per file, text files are capped at 2 million tokens, spreadsheets run about 50 MB, and images 20 MB. Paid plans allow up to 80 file uploads every 3 hours; the free plan allows 3 uploads per day, with a shared 25 GB storage pool.
Here you’re not summarizing a PDF, you’re interrogating it: ask a 60-page lease which clauses auto-renew, cross-check two contract versions, or have it extract every deadline into a table.
Concrete use: upload three supplier quotes as PDFs and ask for a side-by-side comparison of price breaks, lead times, and penalty clauses: a task that used to eat an afternoon.
6. Hand off multi-step tasks to an agent
This is the biggest shift of 2026: assistants that execute work instead of describing it. OpenAI launched ChatGPT Work on July 9, 2026, an agent for “longer, more involved tasks” on paid plans except Free and Go. Its Plan mode gathers context, asks questions, and shows you a step-by-step plan to approve before anything runs. The output is a deliverable: a document, presentation, analysis, or interactive website built with its Sites feature. The standalone agent mode is consolidated into Work.
On September 29, 2026, OpenAI added Dots: “always-on agents” that get their own cloud computer and browser and work toward ongoing goals. Your first dot is included with a Pro or Business Premium plan at no extra cost. Limits: background work is read-only (a dot can’t send messages or change your apps unless permitted), OpenAI says “dots can still make mistakes, so always review consequential work,” and Dots weren’t available in the EEA, Switzerland, or the UK at launch.
Concrete use: brief a dot on your podcast once, and it keeps drafting show notes and social clips from every new episode while you record the next.
7. Put recurring work on a schedule
Assistants can run tasks later, or over and over, without you prompting them. ChatGPT’s scheduled tasks handle one-time reminders, recurring work, and monitoring: “let me know when my package is delivered” is a task, not a question. Since June 2026 there’s a dedicated Scheduled page to manage them, and since August, tasks can trigger from events in Gmail, Slack, and GitHub.
Limits: free and Go accounts get 3 active tasks running no more than once per day; Plus gets 5; Pro and Enterprise get 15. Event-triggered tasks are paid-plan only and can run up to 30 times per hour. Completed tasks still consume usage allowance, since processing happens in the background.
Concrete use: schedule a Monday 8 a.m. digest that searches the web for industry news and delivers five bullets to your phone before your first coffee.
8. Commission a research report with citations
Deep research turns a question into a multi-source report with links you can verify. In ChatGPT, you describe the outcome, review and modify the research plan before it starts, watch progress, and receive a structured report with citations. It only takes read actions, works with uploaded files and supported apps, and availability depends on your plan and country. Free users get a lightweight version, around 5 tasks a month under the limits OpenAI announced in April 2025.
Google runs the same play: Gemini Deep Research drafts a custom research plan you can edit, then browses and delivers a structured report, and it’s reachable inside Gmail and Docs through the Ask Gemini panel.
Concrete use: before a sales call, commission a report on a prospect’s market: you’ll walk in with cited facts instead of what you half-remember from LinkedIn.
9. Build small apps and tools without writing code
Claude can turn a prompt into a small working app that other people can use. Artifacts are the documents, dashboards, and interactive tools Claude creates beside your chat, and they can call Claude directly, becoming small apps. Per Anthropic’s help center: describe what you want, Claude writes the code, and the app runs on Anthropic’s infrastructure: no API keys, no builder cost, and each user’s usage counts against their own plan. Artifacts start private and are shareable; templates export to Word, PDF, PowerPoint, Google Docs, HTML, and MP4. Storage caps at 20 MB per artifact on paid plans.
OpenAI’s answer is Codex in the ChatGPT desktop app, and Sites inside ChatGPT Work for dashboards and prototypes. The trade-off is real: Claude’s apps run for anyone you share them with, while Sites work lives inside your ChatGPT setup.
Concrete use: describe “a tip calculator that splits by percentage and rounds up” and you’ll have a working tool to share before the coffee’s cold.
10. Customize the assistant with Skills, Projects, and plugins
You can teach an assistant your rules once, and it applies them everywhere. Claude Skills are folders of instructions, scripts, and resources that load only when relevant. Anthropic ships built-in ones for Excel, Word, PowerPoint, and PDF creation, and you can upload your own even on the free plan. Claude Projects add a knowledge base: upload documents and set instructions that apply to every chat in that workspace. The Skills format is also an open standard at agentskills.io, so your work isn’t locked to one vendor.
ChatGPT’s route changed: OpenAI announced on September 11, 2026 that custom GPTs retire on December 11, 2026 (February 11, 2027 for some enterprise workspaces): instructions become a skill in a plugin, knowledge files become reference files, but custom API actions don’t transfer. If you still rely on a custom GPT, migrate it now.
Concrete use: put your brand style guide into a Project or Skill once, and every future draft starts on-brand without you re-explaining.
11. Connect Google Drive, GitHub, Gmail, and the rest of your stack
Assistants can now read (and sometimes act) inside the apps you already use. ChatGPT’s connected apps include Google Drive, Slack, SharePoint, Microsoft Teams, Outlook email and calendar, GitHub, GitLab, Box, Notion, and iCloud Mail and Calendar. You can search files (“find my project brief in Drive and summarize next steps”), use apps as deep research sources, or take supported write actions, with approval settings from “Always ask” to “Allow all actions.” Claude offers Google Workspace connectors for Gmail, Calendar, and Drive, plus thousands of MCP-based third-party connectors.
Security note: OpenAI’s docs state app permissions don’t grant access beyond what you approved, but think before connecting work email, on free or personal plans your chats may train the model unless you opt out in Data Controls.
Concrete use: connect Drive and ask “which docs mention the Q3 launch date, and is it consistent?”: a five-minute audit of months of scattered writing.
12. Search the live web with real sources
Assistants answer with current information and show their links, which changes how much you can trust them. ChatGPT search, launched in October 2024, delivers timely answers with clickable sources and can use your approximate location for local results. Gemini is wired into Google Search by default, and ChatGPT’s voice mode can search aloud. Deep research, item 8, is this capability turned up to eleven.
Web search reduces hallucination on current events. It does not eliminate it. Sources are shown so you can click them. Click them when the stakes matter.
Concrete use: ask “what changed in my state’s small-business tax rules this quarter?” and open the cited government pages before you act on the summary.
Which assistant does what best
Based strictly on what each company’s documentation confirms as of November 2026:
| What you want to do | Strongest pick | The detail that decides it |
|---|---|---|
| Multi-step work with a deliverable | ChatGPT (Work) | Plan mode approval, decks/docs/sites output; paid plans except Free and Go |
| Always-on background agent | ChatGPT (Dots) | First dot included with Pro or Business Premium; read-only in background |
| Small apps others can use | Claude (Artifacts) | Runs on Anthropic’s infra, no API keys, viewers’ usage counts against their plan |
| Live camera or screen help | Gemini (Live) | Camera and screen share on Android/iOS; frames processed periodically |
| Spreadsheet cleanup and charts | ChatGPT (data analysis) | Python notebook, interactive bar/line/pie/scatter charts |
| Cited long-form research | ChatGPT or Gemini | Both show an editable plan, then a cited report |
| Reusable custom instructions | Claude (Skills/Projects) | Open standard format; works even on the free plan |
Per ChatGPT’s pricing page (November 2026): Free, Go at $8/month, Plus at $20/month, Pro in $100-$500 tiers. Claude and Gemini both have capable free tiers. Artifacts and Skills cost nothing.
Where to start
Pick one feature that removes real friction from your week and try it on a free plan today: uploading a spreadsheet for analysis is the highest-value first move. If you hit usage limits, ChatGPT Plus at $20/month is the sensible first upgrade. Add Claude if app-building or branded documents are your bottleneck; use Gemini if your days run on Android and Google Workspace. Once a month, review your assistant’s memories and connected-app permissions: the convenience is real, but so is the access you’re granting.
FAQ
What can AI assistants do besides answering questions?
As of November 2026, ChatGPT, Claude, and Gemini can generate and edit images, hold voice conversations, take camera or screen input, analyze spreadsheets and build charts, run tasks as agents (ChatGPT Work and Dots), produce cited deep research reports, build small apps (Claude Artifacts), connect to Google Drive, Gmail, GitHub, and Slack, and search the web with source links. Advanced features require paid plans.
How does ChatGPT Work’s agent mode work?
You describe the outcome, and Plan mode gathers context, asks clarifying questions, and produces a step-by-step plan you approve before work starts. The agent then executes on OpenAI’s servers using connected apps and returns finished deliverables: documents, presentations, analyses, or websites built with Sites. It launched July 9, 2026 on paid plans except Free and Go.
How much do these AI assistant features cost?
Per ChatGPT’s pricing page (November 2026): Free is $0, Go $8/month, Plus $20/month, Pro $100, $200, or $500. Images, web search, voice, file uploads (3 per day), and lightweight deep research work free. ChatGPT Work and event-triggered tasks skip Free and Go; Dots require Pro or Business Premium. Claude’s Artifacts and Skills work even on its free plan.
What are the main alternatives to ChatGPT for these features?
Claude leads for small apps (Artifacts) and reusable Skills; Gemini for live camera and screen input and Google Workspace integration; Perplexity for cited search. Microsoft’s Copilot embeds similar abilities into Windows and Office. Most people pick one primary assistant and one specialist. ChatGPT for agents plus Claude for app building, say.
What are the main limitations of AI assistants?
Agent actions require approval and can still be fooled by malicious web pages (OpenAI’s own warning); the data-analysis sandbox can’t fetch external data; scanned PDFs often produce unreliable values; ChatGPT’s video and screen sharing is confined to Advanced voice mode on mobile; Dots were unavailable in the EEA, Switzerland, and the UK at launch; and custom GPTs disappear on December 11, 2026.
Can I use AI assistant output commercially?
Generally yes. OpenAI’s terms say you own the output you generate, and Anthropic’s consumer terms assign its rights in outputs to you, though AI-generated work may not be copyrightable, and you must follow each provider’s usage policies. Never paste confidential client data into consumer-plan assistants, and get professional advice for compliance-sensitive work.
How can users get better results from AI assistants?
Give the assistant your goal, constraints, and an example of good output; upload the real file instead of describing it; use Plan mode or deep research’s editable plan to steer big tasks before they run; and specify the output format. For repeated work, save instructions in a Project, a memory, or a Skill, and spot-check citations.
What should users choose or do next?
Pick the two features that would save you the most time and try them on a free plan this week: likely uploading a spreadsheet for chart analysis, or setting one scheduled task like a Monday news digest. If you routinely need finished deliverables, ChatGPT Plus ($20/month) is the best upgrade; add Claude for small apps, or Gemini if you live in Google Workspace.
Sources
- ChatGPT Release Notes: OpenAI Help Center
- ChatGPT Work: OpenAI
- Introducing dots: OpenAI
- ChatGPT agent: OpenAI Help Center
- Scheduled tasks in ChatGPT: OpenAI Help Center
- Deep research in ChatGPT: OpenAI Help Center
- Data analysis with ChatGPT: OpenAI Help Center
- Uploading files and audio to ChatGPT: OpenAI Help Center
- Images in ChatGPT: OpenAI Help Center
- ChatGPT Voice. OpenAI Help Center
- Memory in ChatGPT: OpenAI Help Center
- Connected apps in ChatGPT: OpenAI Help Center
- Custom GPT retirement and migration FAQ: OpenAI Help Center
- Searching the web with ChatGPT: OpenAI Help Center
- Introducing ChatGPT search: OpenAI
- Terms of Use. OpenAI
- Consumer Terms of Service: Anthropic
- ChatGPT pricing: OpenAI
- What are artifacts and how do I use them?: Claude Help Center
- What are skills?: Claude Help Center
- What are projects?: Claude Help Center
- Use Google Workspace connectors. Claude Help Center
- Gemini Deep Research: Google blog
- Meet Deep Research: Google Workspace blog
- Talk naturally with Gemini Live. Google Support
- Gemini Live camera and screen sharing: Google Store
Weekly digest
Get our weekly AI digest
The latest AI tools, prompts, and insights — short, useful, every week.
AIUnpacker Editorial
Articles are written for a specific question. Sponsored work is labeled.
