ChatGPT's @Presentations Skill: What It Can and Can't Do (2026)

ChatGPT's @Presentations skill turns attached source material — notes, research, data, or an existing deck — into a Google Slides or PowerPoint file from inside a ChatGPT conversation. You invoke it with

, optionally point it at a template to match, and it plans the slides, builds the file, and validates the rendered output. It's the most accessible way to get a real deck out of ChatGPT — and OpenAI's own documentation is upfront that template matching and advanced formatting "may require manual refinement." This review covers the workflow, the genuine strengths, and the boundaries where you need different tooling.

What is the @Presentations skill?

@Presentations is a first-party ChatGPT capability (documented in OpenAI's ChatGPT use-case guides) that creates, edits, and previews slide files without leaving the chat. It works alongside two companions: the Google Drive connector, which reads source files and can update native Google Slides directly, and a Template Creator that converts an existing PowerPoint into a reusable template. Availability varies by plan and workspace configuration — it's aimed squarely at ChatGPT Work users.

OpenAI's starter prompt captures the intended workflow in one line:

"Use @Presentations to turn the material I've attached into a presentation for its intended audience. If I've included an existing deck or template, match its structure, visual style, and branding."

The workflow, tested against the docs

1. Attach source material and name the audience. The skill is material-first — it compresses what you give it rather than inventing content, which is the right default (the same principle we push in how to make a presentation with AI).
2. Optionally reference a deck or template. ChatGPT attempts to match its structure, style, and branding.
3. ChatGPT plans, builds, validates. It produces a slide plan, generates the file, and checks the rendered slides before handing them over.
4. Review and refine. Formatting fixes and template deviations get resolved manually or through follow-up requests.

What it's genuinely good at

• Zero setup. No API keys, no config files, no installs. For a ChatGPT Work user, it's the shortest path from "notes" to "deck file" that exists.
• Drive-native editing. Updating an actual Google Slides file in place — not an export — is something most third-party tools can't do.
• Material-grounded drafts. Because the workflow starts from attachments, output quality tracks input quality, which beats topic-only generation every time.
• Template Creator. Converting a house deck into a reusable template is a real workflow feature, not a demo.

The documented limits (and what they mean in practice)

OpenAI's guide lists the caveats itself, which we'll take at face value plus practical translation:

To these, add the structural boundaries no chat-app skill can cross: there is no API surface — you can't script it, schedule it, or call it from a pipeline; no repo or database context — it sees what you attach, not your codebase or warehouse; and no batch mode — fifty per-client variants of one deck means fifty conversations.

When @Presentations is enough — and when it isn't

It's enough when: you're a ChatGPT Work user making occasional decks from documents, your template tolerance is "close is fine," and the deck's final home is Google Slides.

It isn't when:

• Decks recur or ship at scale. Weekly reports and per-customer decks belong in an API workflow — programmatic generation, template-locked output, batchable.
• Your agent should do it as part of a larger job. Coding agents (Codex, Claude Code, OpenClaw, Cursor) reach slide generation through agent skills or an MCP server — with repo context and chaining that a chat app can't offer.
• You need beyond-deck output. Narrated presentation videos, 4K image-designed slides, or 19-language batch localization are pipeline features — see the narrated video guide.
• Brand exactness is non-negotiable. Template-fill systems enforce layout by construction; style matching approximates it.

The honest framing: @Presentations competes with "making the deck yourself," and wins. It doesn't compete with production pipelines, and OpenAI doesn't claim it does.

FAQ

Is the @Presentations skill free?

It ships with paid ChatGPT plans, with availability tied to plan tier and workspace configuration — OpenAI notes feature access varies. There's no separate per-deck fee beyond your subscription.

Can @Presentations output real PowerPoint files?

Yes — it creates and edits both Google Slides and PowerPoint files, and previews rendered slides in-chat. Post-export formatting fidelity is where OpenAI's "manual refinement" caveat applies most.

Can it match my company's slide template?

It will attempt to match an attached deck's structure, visual style, and branding, and the Template Creator can turn a house deck into a reusable template. Exact brand compliance still typically needs a manual pass — or a template-fill system where the layout is enforced rather than imitated.

How is this different from a Codex or Claude Code PPT skill?

Same "skill" vocabulary, different worlds: @Presentations lives in the ChatGPT app for interactive use; coding-agent skills are installable packages that run with repo context, scriptability, and chaining. The Codex skills guide covers that side of the fence.

Can I automate @Presentations with the OpenAI API?

No — it's a ChatGPT app capability, not an API endpoint. For programmatic slide generation you need a presentation API; our developer comparison with code examples walks through the options.