# How to Add Your Help Center Content to Intercom Fin

*How-to — 2026-10-15 — by Mahmoud Zalt*

A practical walkthrough of Fin AI Agent, Train, Content: public articles, snippets, internal articles, PDFs, website sync and external sources.

**Short answer.** **Everything Fin can read lives in one place: Fin AI Agent, then Train, then Content.** From that screen you add public articles, snippets, internal articles, uploaded documents, a website sync, and connections to outside knowledge tools. Adding a source is not enough on its own, each piece also has to be enabled for Fin AI Agent, which is the step most people miss on their first pass.

**About this guide.** This guide is based on publicly available information and Intercom's own documentation, checked on 17 August 2026. Sistava is not affiliated with Intercom. Interfaces and pricing change often, so confirm the current steps in their documentation before you rely on them.

Fin is good. That is worth saying up front, because the most common reason a Fin trial disappoints is not the model, it is the reading material. Fin answers from the content you point it at, and if that content is thin, stale, or never got switched on, the answers will be too.

This guide walks the content side of Fin in the order that actually works. Where the screen is, what each source type is for, how to enable each one, and the quiet gotchas that cost people a week.

Do this before you touch guidance, procedures or workflows. Content is the layer everything else sits on, and tuning tone on top of a half empty knowledge base is wasted effort.

## Where is Fin's content screen?

Intercom's documentation gives one path: Fin AI Agent, then Train, then Content. That single screen is where you add sources, manage them, and preview what Fin is doing with them.

Intercom describes this section as the place to add, manage and preview the sources Fin learns from, and frames it as somewhere you come back to periodically rather than a one time setup. That framing is accurate. Content maintenance is the ongoing job, not the install.

## Which content types can Fin actually use?

Intercom publishes a compatibility table showing which sources power which surface. The short version: public articles are the only source that also appears in your customer facing Help Center. Everything else is invisible to customers browsing your docs, but readable by Fin.

| Source | Fin AI Agent | Copilot | Help Center |
|---|---|---|---|
| Public articles | Yes | Yes | Yes |
| Internal articles | Yes | Yes | No |
| Snippets | Yes | Yes | No |
| Websites | Yes | Yes | No |
| PDFs | Yes | Yes | No |
| Zendesk, Confluence, Guru, Notion | Yes | Yes | No |
| Macros | No | Yes | No |
| Conversation history | No | Yes | No |

That table settles a question people ask constantly. Yes, Fin can answer from material you never publish. Internal articles and snippets exist precisely for the policy details you would rather not put on a public docs page.

## How do you add each source type?

Every route starts from the same place. Open Fin AI Agent, then Train, then Content, and work from the Add content section. The less obvious sources sit behind See all.

### Adding content to Fin

1. **Public article** — Choose Public article under Add content. Write it, publish it so its status shows as published, and turn on the toggle for Fin AI Agent. Intercom notes an article does not have to be added to a Help Center collection for Fin to use it.
2. **Snippet** — Choose Snippet under Add content. Snippets are short pieces such as a single FAQ or a time sensitive notice. Enable the Fin AI Agent toggle in the Details panel or Fin will not read it.
3. **Internal article** — Choose Internal article under Add content. In the Details panel you can enable it for Fin AI Agent, for Copilot, or both. Internal articles never appear in your public Help Center.
4. **Website sync** — Choose Website sync under Add content and enter public URLs, for example your pricing page or your blog. Intercom's documentation says synced sites refresh weekly, and the content becomes usable once the first sync completes.
5. **Upload a document** — Select See all under Add content, then Upload a document. Intercom's docs describe PDF upload with the text extracted within minutes. Confirm the current accepted file types in their documentation before you plan a bulk upload.
6. **External knowledge source** — Select See all, then pick the source. Intercom's page lists Zendesk, Guru, Notion and Confluence with an import or sync option, and Salesforce, Box, Freshdesk and Document360 as sync only.

**The toggle is the whole game.** Three of these paths end with a switch you have to turn on, and the article, snippet or internal article looks completely finished without it. If Fin is confidently telling customers it does not know something you documented last week, check that toggle before you check anything else.

## How much content does Fin need before it can go live?

For chat there is a hard floor. Intercom's deployment documentation states you need at least ten public articles live in your Help Center before you can set Fin live over chat. That is a gate, not a suggestion.

Ten is the minimum that lets you press the button, not the amount that produces good answers. Treat it as the moment you are allowed to start testing, and expect the real number for a decent resolution rate to be considerably higher.

Volume is the easy part. Quality is what separates a Fin deployment that resolves half the queue from one that mostly escalates, and quality is entirely under your control.

## How should you write articles so Fin can use them?

Write one question per article and put the direct answer in the first paragraph. That single habit does more for an AI agent's accuracy than any setting, because it gives the retrieval step an unambiguous match instead of a page it has to interpret.

Independent practitioners writing about Fin resolution rates keep landing on the same list, and it matches what Intercom says about content quality. The specifics worth copying:

- **Answer first. **Lead with the answer, then add the context. Do not build up to it.
- **One topic per article. **A page covering six loosely related things matches all six queries weakly.
- **Specific titles. **A page called Getting Started tells a retrieval system almost nothing about what is inside it.
- **No instructions that only exist in a screenshot. **Intercom's own FAQ notes Fin cannot read the alt text in images, so anything shown only as a picture is invisible to it.
- **Remove duplicates. **Two articles saying slightly different things about the same policy is a coin flip on which one Fin quotes.
- **Refresh the stale ones. **Articles nobody has touched since a product rename are the most reliable source of confidently wrong answers.

## What are the limits and gotchas?

A few things are stated plainly in Intercom's documentation and are worth knowing before you architect around them, because each one has quietly broken somebody's plan.

- **External sources are capped. **Intercom's Fin FAQ page states up to 100 external content sources.
- **Website sync is weekly, not live. **If you change a pricing page today, do not assume Fin knows today.
- **Images are not read. **Per Intercom's FAQ, Fin cannot read alt text in images.
- **Some reasoning is out of scope. **The same FAQ page notes Fin cannot reliably calculate dates or business days, and cannot book, schedule or confirm meetings.
- **Language coverage is broad but bounded. **Intercom lists 45 plus supported languages and says behaviour outside that list is unpredictable.

None of these are flaws exactly. They are the shape of a retrieval based agent, and knowing the shape is what stops you from writing a delivery date article that Fin cannot reason about anyway.

## There is another way to think about this

Notice what the work in this article actually is. You are assembling a reading list for a machine, then flipping switches to tell it which items on the list are allowed, then maintaining that list forever as your product changes.

That plumbing is exactly what we removed at Sistava. Rather than curating sources and toggling each one, you hire an AI Employee and tell it, in plain English, what you want handled. It has its tools connected already, it asks you when it is unsure, and it sets itself up. There is no content pipeline to wire and nothing to keep in sync.

That is a genuinely different trade, not a better version of the same one. If you already have a mature Help Center and a team that maintains it, Fin's model turns that investment into answers immediately, and that is a real advantage. If you do not have those docs yet, building them purely to feed a bot is a large project hiding behind a small setup screen.

Either way, finish the content work properly. A support agent that reads bad documentation gives bad answers, and that is true of every vendor in this category including ours.

## How do you check what Fin is actually using?

Use the preview panel and the reporting content table together. Intercom's documentation describes previews as available throughout Fin's training areas, including Content, so you can type a question and watch which source it pulls without touching a real customer.

Then, once Fin is handling live traffic, the Fin AI Agent report includes a content performance table that can be sorted by involved conversations or by resolved conversations. That table is the honest answer to which articles are earning their place and which ones nobody has ever needed.

## FAQ

### Where do I add content to Intercom Fin?

Go to Fin AI Agent, then Train, then Content. That screen holds every source type, with the less common ones behind the See all option in the Add content section.

### How many articles does Fin need before it can go live over chat?

Intercom's chat deployment documentation states you need at least ten public articles live in your Help Center to set Fin live over chat. That is the minimum required to deploy, not the amount needed for good coverage.

### Can Fin use content that is not published in my Help Center?

Yes. Internal articles and snippets are readable by Fin and by Copilot but never appear in the customer facing Help Center, according to Intercom's knowledge sources documentation.

### Can Fin read PDFs?

Intercom's documentation describes uploading a document from Fin AI Agent, Train, Content by selecting See all and then Upload a document, with PDF text extracted within minutes. Check their current documentation for the accepted file types before planning a bulk upload.

### How often does website sync refresh?

Intercom's documentation states synced website content refreshes weekly. If you change a public page today, do not assume Fin has picked it up today.

### Why is Fin ignoring an article I published?

The most common cause is the Fin AI Agent toggle. Public articles, snippets and internal articles all need to be explicitly enabled for Fin, and they look complete without it.

### Is there a limit on external content sources?

Intercom's Fin FAQ page states a limit of up to 100 external content sources. Confirm the current figure in their documentation, since limits change.

## Sources

- [Intercom Help, ](https://www.intercom.com/help/en/articles/7837514-add-your-content-for-fin-ai-agent)Add your content for Fin AI Agent
- [Intercom Help, ](https://www.intercom.com/help/en/articles/9440354-knowledge-sources-to-power-ai-agents-and-self-serve-support)Knowledge sources to power AI, agents and self-serve support
- [Intercom Help, ](https://www.intercom.com/help/en/articles/8286630-deploy-fin-ai-agent-over-chat)Deploy Fin AI Agent over chat
- [Intercom Help, ](https://www.intercom.com/help/en/articles/7837535-fin-ai-agent-faqs)Fin AI Agent FAQs
- [Intercom Help, ](https://www.intercom.com/help/en/articles/7837533-fin-ai-agent-reporting)Fin AI Agent reporting
- [Fin by Intercom Help Center, ](https://fin.ai/help/en/articles/11378444-adding-content-to-fin)Adding content to Fin
- [Pageloop, ](https://www.pageloop.ai/blog/how-to-improve-intercom-fin-resolution-rate)How to improve Intercom Fin resolution rate

**Tags:** intercom-fin, intercom-tutorial, ai-customer-support, help-center, knowledge-base