> For the complete documentation index, see [llms.txt](https://docs.papers.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.papers.ai/using-ai/using-ai.md).

# Using AI

Ask AI questions about your project, get it to draft or edit content, and understand exactly what it can see.

The AI panel sits in the right sidebar. It reads the project you have open, accepts files you attach, and can propose edits to your documents.

## Send a prompt

{% stepper %}
{% step %}

### Open the AI tab

Open the right sidebar and switch to the **AI** tab.
{% endstep %}

{% step %}

### Pick a model

Pick a model from the model selector. See [Adding and managing models](/using-ai/ai-models-and-your-own-model.md) if none are enabled yet.
{% endstep %}

{% step %}

### Send your prompt

Type your question or request and send it. Attach files or images when you want the AI to use their contents.
{% endstep %}

{% step %}

### Review the result

Review the response, and any proposed edits. See [Reviewing AI changes](/using-ai/reviewing-ai-changes.md).
{% endstep %}
{% endstepper %}

{% hint style="success" %}
The AI is useful well beyond drafting text. Ask it to summarize a dataset, explain a compile error, draft a bibliography entry, or turn a rough note into a structured section. It works across every file type in your project, not just prose.
{% endhint %}

## What the AI can see, and what it can't

Getting a good answer usually comes down to knowing what's actually in context:

* **The file you have open**: its path, so the AI knows what "this file" means.
* **Your project's file tree**: every file's name and path, but not its contents.
* **Files you explicitly attach** to a message: their full contents, for that conversation only.
* **Files the AI opens itself** while working on a request, if the model supports that.
* **Earlier messages in the current conversation**: a new conversation starts with no memory of past ones.

It does **not** automatically see: files in other projects, anything outside Papers AI, comment threads, version history, or your Kanban or timeline boards. If you want the AI to act on a comment or a task, paste the relevant text into your prompt.

{% hint style="success" %}
If an answer feels vague or the AI seems to be guessing, it's almost always a missing attachment. Attach the specific file it needs: three small, relevant attachments beat asking it to "look at the project."
{% endhint %}

## Privacy

The AI sends whatever it can see to the model you've selected. If you're using a cloud model, your prompt and attachments go to that provider. If you'd rather keep everything on-device, connect a local model. See [Adding and managing models](/using-ai/ai-models-and-your-own-model.md).

## Troubleshooting

<details>

<summary>No response, or it stalls</summary>

Use stop or retry, and check your connection.

</details>

<details>

<summary>It only chats and won't make edits</summary>

The selected model doesn't support actions (chat-only mode). Switch models. See [Adding and managing models](/using-ai/ai-models-and-your-own-model.md).

</details>

<details>

<summary>Attachments won't attach</summary>

Confirm the file type is supported and retry.

</details>

## Next steps

* [Reviewing AI changes](/using-ai/reviewing-ai-changes.md)
* [Adding and managing models](/using-ai/ai-models-and-your-own-model.md)