> ## Documentation Index
> Fetch the complete documentation index at: https://docs.attensira.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Wins: the first time an answer names you

> What counts as a win, how the hourly pass decides it, the message you get, the share link, and the win_detected event.

## What a win is

A **win** is the first time an engine names your brand, or links to your own site, in an answer to a prompt it had been answering without you. It is recorded per prompt and per engine, so ChatGPT naming you on a prompt and Perplexity naming you on the same prompt are two wins.

There are two kinds, and each is its own first:

| Kind | What happened in the answer |
| - | - |
| **Named** | The answer named your brand in its own words |
| **Linked** | The answer cited a page on your domain, or on a subdomain of it |

An engine can name a brand it never links, and link a page while recommending somebody else, so one does not stand in for the other. Each (prompt, engine, kind) can be a win **once**, ever.

A win says what happened and when we started reading. It never says why. The message reads "for the first time since we started reading this prompt", not "after we shipped" or "because of our work".

## What counts

Every hour, a pass reads the last three days of your [readings](/measure/how-we-collect-data), in your workspace's time zone. It is arithmetic over the same numbers your dashboard shows, with no model involved, so the same readings always produce the same wins.

A naming is a win only when all of these are true:

<Steps>
  <Step title="The prompt does not name your brand">
    A prompt that names you ("is Acme any good") is a question about you, and being named in its answer is not news. The prompt's text is checked against your brand's name, your domain and your domain's bare name, as whole words. A prompt that names a **competitor** ("Acme vs Curve") still counts: that is a question a buyer asks.
  </Step>

  <Step title="You still track it, on that engine">
    A prompt you deleted, or an engine you removed from a prompt, produces no wins.
  </Step>

  <Step title="We had read it for at least three earlier days">
    Before the day you were named, at least **three** earlier days carried at least one successful reading of that prompt on that engine.
  </Step>

  <Step title="No earlier day had you in it">
    On no earlier day with a successful reading did the answer name you (for a named win) or link to you (for a linked win). Every earlier day is checked, not only the last three.
  </Step>
</Steps>

Answers for the same prompt from different surfaces or countries on the same day are one day here. The question is only whether you were in the answer.

### Failed reads never count

A day counts only if at least one reading of the prompt on that engine succeeded. A day of failed reads and a day with no reading at all are the same to this pass: silence. A failed day cannot be one of the three earlier days, and it cannot be the day you were named.

### What is never announced

* **Names from before detection started.** The first pass on a workspace records everything you are already named or linked for as a baseline and announces nothing. Turning detection on does not announce months-old news as a first.
* **A naming we could not call a first.** If an engine named you on a prompt before we had three successful earlier days of reading it, or had named you on it before, that (prompt, engine, kind) is recorded as a baseline. It is never announced and never becomes a win later.

Baselines are never listed, shared or counted as wins.

## The win message

When the pass finds new wins, you get **at most one message per workspace per day** in your workspace's time zone, listing every win not told yet. A win found after that day's message went out is in the next day's message.

Each win is told in one sentence:

> ChatGPT named Acme for the first time since we started reading this prompt on 2 September 2026, in an answer to 'best dental software' on 28 September 2026.

"Since we started reading" is the first day we have a successful reading of that prompt on that engine. A linked win says "linked to acme.com" instead of "named Acme".

Under the sentence:

* **The quote.** The engine's own words, verbatim, cut from the answer we read: at most 400 characters around the first mention of your brand (or the cited link, for a linked win). When the quote is cut from a longer answer, it is shown with "..." at both ends. A quote is never edited or summarised. If no answer that day yields a quote that contains the mention and passes our honesty check, the win has no quote rather than a tidied one.
* **Related work**, only when it is true. If a piece of work that lists this prompt went live in the 90 days before the win, the message adds "Work on this prompt: *title*, live since *date*." as a separate line. It does not claim the work caused the win, and work that is not live yet is never mentioned.

### Where it goes

The message goes by email by default. Under **Settings → Notifications → Wins** you can send it to Slack instead, or to both; Slack posts to the same channel as the Monday note. See [Notifications](/account/notifications).

The win email is its own notification category, so you can unsubscribe from it without turning off anything else.

## The win card in the dashboard

The newest win nobody in your workspace has seen yet appears as a card at the top of **Home**, and at the end of every project's feed. Most visits have none, and then nothing is drawn.

* **Shown once.** The card counts as seen as soon as it is on screen, so it does not come back on your next visit, on Home or in a project.
* **Share in one tap.** The card carries LinkedIn, Slack and **Copy link**. Slack posts the card to the workspace's Slack channel; see [Sharing a win card](#sharing-a-win-card) for what the link is.
* **Close** takes the card off the screen. The win stays in your list of wins, and its link, if it has one, keeps working. **Dismiss** puts the win away for good and, when it has a link, turns that link off in the same step. The button says **Dismiss and turn off the link** when there is one.

### The referral offer

After you share a win, and in the card's place when you close or dismiss it, you may see **Give a month, get a month**: your referral link to send to a team that should be in AI answers too. When they start a plan with it, you each get a month. It appears only after a win, and only when your organisation is eligible; **Not now** hides it for the rest of that visit. See [Referrals](/account/referrals) for who qualifies, the link and the rules.

## Sharing a win card

A win card shows the quote, the engine, the prompt and the date. Sharing it to LinkedIn or copying its link makes a **public link** to the card.

* **One win, one link.** The first share makes the link and every later share reuses it, so a LinkedIn post and a copied link open the same card. On LinkedIn, the link is opened in LinkedIn's own share box; Attensira does not post anything for you.
* **Permanent until you turn it off.** The link has no expiry of its own. It stops working when you **unshare** the win or **dismiss** it. Dismissing puts the win away and turns off its link in the same step.
* **A turned-off link stays off.** Sharing the win again makes a new link. The old one never works again.
* A dismissed win cannot be shared.

### What a stranger sees

The person who opens the link needs no account. The page shows only your brand, your domain, your logo (when your brand profile has one), the engine, the prompt, the quote and the date. It shows no ids, no counts, nothing else from your workspace, and no Attensira branding, so an agency can share a client's card as it is.

A link that was turned off, or was never a win card, shows "This card is not available".

## The `win_detected` event

Every win is also recorded as one `win_detected` event, with the win's id as its subject. Its payload carries:

| Field | What it is |
| - | - |
| `win_id` | The win |
| `prompt_id`, `prompt_text` | The prompt |
| `platform` | The engine |
| `kind` | `named` or `linked` |
| `slot_day` | The day of the answer that named you, in your workspace's time zone |

Today no automation you create reacts to it. `win_detected` is not one of the event types you can pick in the dashboard, over MCP or over REST, and each of those asks for at least one type, so a new [event automation](/agent/autopilot#event-automations) never listens for it. The built-in **Event responder** does not listen for it either. Only an automation saved with an empty event list would, because an empty list means every event type.

## Reading your wins

* **Over MCP**, the `list_wins` [tool](/mcp/tools) returns your wins newest first, each with the same sentence the message uses, the quote, the date, "since we started reading" and the share link when there is one.
* **In the [morning brief](/agent/morning-brief)**, the Monday edition counts the week's wins.
* **In the dashboard**, the newest unseen win is a card on Home and in project feeds. See [The win card in the dashboard](#the-win-card-in-the-dashboard).

Dismissed wins are left out of the list.

## Who has it

Wins run on every workspace with hourly [detection](/agent/findings#who-has-it). Paid workspaces are enrolled automatically, the first time their subscription is active.

If the pass fails for a workspace, the failure shows on [loop health](/agent/loop-health) as **First-naming detector**, and the next hour's pass tries again. If the day's message could not be queued on any channel you chose, its wins stay stored and untold, and the next hour's pass tries again; a win is never lost.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.