> For the complete documentation index, see [llms.txt](https://help.rankability.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.rankability.com/serena/advisor-knowledge-base-and-data-connections.md).

# Serena sources and data connections

Ground Serena with durable client sources and correctly scoped integrations while preserving processing, freshness, and privacy boundaries.

Serena produces better analysis and execution when the selected client contains reliable evidence. Sources and integrations both add context, but they have different permissions, freshness rules, and failure modes.

## Know the difference

* **Sources** are durable client reference material. Supported options include uploaded files, pasted text, website or YouTube URLs, multiple pasted URLs, website imports, and Google Drive files when that option is configured.
* **Integrations** authorize Rankability to read data or deliver supported work through a service, property, profile, repository, publishing destination, or task system.
* **Chat attachments** ground a conversation but do not automatically become durable client sources.

Adding a source does not connect the underlying service account. Connecting an account does not import every file, property, or page into Serena's client context.

## Add useful sources

1. Confirm the selected client.
2. Open **Settings → Sources** and choose **Add source**.
3. Choose the narrowest useful option shown:
   * **Upload file** for supported documents or structured files.
   * **Paste text** for approved notes, policies, interviews, testimonials, or brand facts.
   * **Website or YouTube URL** for one page or a video transcript.
   * **Paste URLs** for a reviewed page list.
   * **Import from website** to discover and select a broader set of pages. Enter the website, or a specific sitemap URL when you want Rankability to read that sitemap instead of looking for one.
   * **Google Drive** to select supported files when the workspace offers that integration.
4. Give pasted or uploaded material a recognizable name.
5. Review the displayed usage impact when an import requires crawling or extraction, then confirm the action.
6. Wait until processing finishes before relying on the source.

Start with decision-grade material: current products and services, audience, differentiators, approved claims, policies, case studies, customer language, and writing guidance. Avoid adding many near-duplicate pages merely to increase volume; conflicting or outdated material makes execution less reliable.

Website sources can show monitoring, crawl, change, or failure states. Reprocess a source when the page materially changes. Disable or remove content that should no longer influence future work. An imported Google Drive file is a copy of supported content; do not assume it updates whenever the original file changes unless Rankability shows a refresh.

## Review what Serena remembers

**Workspace notes** in the same panel holds the durable facts and decisions Serena carries into its work for this client — what the business is, its goals and constraints, its preferences, decisions you have made, and lessons from earlier work. Open it to see the current notes, switch to **Archived** for notes no longer in use, and use **Add note** to record one durable fact at a time under the category it belongs to. A note you save yourself counts as confirmed and stays protected until you change it.

When Serena learns something in a conversation that conflicts with a note you already have, the note shows a proposed update for you to review. Serena keeps using the existing fact until you choose **Accept update**; **Keep existing** dismisses the proposal. Where a note came from a conversation, you can open that conversation from the note.

Archiving a note stops Serena using it but keeps it in history, and archived notes can be restored. **Note history** lists recorded changes, oldest entries paging behind **Older changes**; history begins when version tracking was added, so earlier changes may not appear. **Edit Markdown** still opens the whole set as one document — one fact per bullet, and removing a bullet archives that fact rather than erasing it.

## Keep monitored pages current

Rankability can watch important client pages and tell you when their content changes, rather than leaving you to notice. Open **Automate**, choose the pages to monitor, and turn the automation on; you need at least one selected page before it will start. You can also add a page to the knowledge base and begin monitoring it in the same step.

Choose weekly or monthly monitoring, and use **Check now** when you need an immediate comparison. The automation checks the selected pages without starting unrelated client work.

When a check finds a difference, the source is flagged for review. Open **Review changes** to compare the saved content with the live page side by side, then either update the knowledge base with the new content or dismiss the change to keep what you have. Nothing is rewritten until you choose to update, so a detected change is a prompt to review rather than an edit Rankability has already made.

If a page has permanently moved within the same site, Rankability presents the original and destination URLs as a separate **Page moved** review. Approve the new destination only when it is the correct replacement, or keep the original source. Temporary and cross-site redirects do not silently replace a saved source.

## Review the generated entity record

When Rankability builds a client's knowledge base, it generates an **Entity Record** source for the client's identity facts — official name, entity type, canonical website, brand aliases, former names, common misspellings, founding details, key people, headquarters, locations and service areas, contact details, products and services, related brands, official profiles, a short description, and disambiguation notes.

The entity record starts as **Needs human review**. While it carries that state you cannot select or star it, and Rankability keeps it out of content production and research. Open the source, correct anything inaccurate, then choose **Mark reviewed** from its menu. The badge changes to **Reviewed** and the record becomes available like any other selected source.

## Upload PDF and Word sources

Rankability extracts text from text-based PDF and `.docx` files. A document needs at least about 50 extracted characters to become usable grounding. Scanned or image-only PDFs, protected files, corrupt documents, and nearly empty files can fail extraction; legacy binary `.doc` files should be resaved as `.docx` or a text-based PDF.

Do not assume an upload is ready merely because the file was accepted. Wait for processing to finish and confirm the source is available before asking Serena to rely on it. A failed source displays a failure state or re-upload banner. Previously failed PDFs and Word documents that do not have a retained original must be uploaded again; there is no automatic backfill for those old failures.

## Connect performance data and destinations

1. Open **Settings → Integrations** for the client.
2. Choose the required service.
3. Authorize an account with access to the intended property or destination.
4. Select the correct GSC site, GA4 property, GBP location, repository, CMS, or other resource when prompted.
5. Wait for the initial sync and check the connection status before interpreting missing data.

GSC provides search-query and page performance, GA4 provides site behavior and conversions, and GBP provides Business Profile evidence and authorized actions. They are not interchangeable. For the agency authorization flow, read [Connecting client Google services](/account-and-settings/connecting-client-google-services.md).

## How Serena uses the evidence

Serena can combine client identity and brand guidance, processed sources, workspace activity, tool outputs, and eligible connected data. The exact evidence available depends on the current client, source status, selected properties, date ranges, monitoring settings, and the Serena surface you are using.

Treat a recommendation as grounded only when the supporting source or dataset is available and current. A missing metric is not automatically zero. A successful connection badge confirms saved authorization, but it does not prove that the expected property was selected or that its newest data has finished syncing.

## Privacy and client boundaries

* Confirm the client before uploading private material or authorizing an account.
* Use the least-privileged account and API scope that supports the workflow.
* Keep credentials, secrets, and unnecessary legal, medical, financial, or personnel data out of sources.
* Remove former vendor access and obsolete sources promptly.
* Do not use one client's material as evidence for another client.

## Troubleshooting

* **A source remains processing** — Wait, refresh the source list, and check for a failure state before importing it again.
* **A PDF or Word document failed** — For a PDF, confirm it contains selectable text rather than only page images. Remove password protection, repair a corrupt file, or resave it as `.docx`, `.txt`, `.md`, or a text-based PDF, then upload it again.
* **A page cannot be read** — Paste approved text or upload a supported file instead of repeatedly crawling the blocked page.
* **A property is missing** — Reauthorize with an account that has access and finish property selection.
* **Data is empty after connection** — Verify the property, date range, sync status, and required monitoring capability.
* **Serena used old information** — Update, reprocess, disable, or remove the stale source, then start a new analysis with the corrected evidence.

For organization apps, Slack, Drive behavior, and external OAuth clients, continue with [Using Serena with connected apps](/serena/serena-integrations.md). For the chat workflow itself, see [Using Serena chat](/serena/using-advisor.md).