# Win a buyer question you are losing

Take one question where AI assistants do not recommend you, and find out why and who they recommend instead. Hand the fix to the right agent, then run the collection again to see whether the answers moved.

The example follows Acme, which sells a payments API. Its report shows that when buyers ask "Which payments API handles split payouts for marketplaces?", the assistants recommend Hooli.

## What you need

- A completed Buyer questions report. To get your first one, see [Get your first AI report](/get-started/first-ai-report/).
- Your competitors in your Company Profile, so the report can show who was recommended instead. See [Company Profile and Domains](/agent/company-profile/).
- For a docs fix: the Docs Agent, with the repository that holds your docs connected. See [Connect GitHub](/agent/github-app-setup/) or [Connect GitLab](/agent/gitlab-setup/).
- For a post: the Marketing Agent.

## 1. Find a question you lose

Open the report and scroll to **Every buyer question**. Questions are grouped by outcome, with the questions you lose first. A question is marked **You lose** when 20% or fewer of its answers passed.

Click Acme's split payouts question to open it. **Checks, worst first** lists the question's checks, and **An answer that did not pass** shows one of the answers, with **Read the full answer** for the whole text. For what every part of the report means, see [Read a Buyer questions report](/journeys/read-a-report/).

## 2. See who was named instead

The question's **Recommended instead** column, and **Recommended instead of you** in the report's headline card, show the competitors named in the answers you did not win. Acme sees Hooli there.

Under **Also named, not on your list**, the report lists other products the answers put forward. Click **Add to competitors** to add one to the competitors in your Company Profile.

## 3. Hand the fix to an agent

In the open question, **What went wrong** lists the causes the report found, each with a button. You can also start from **Fix these first** near the top of the report, where each card shows how many buyer questions it affects and has **Start the fix**.

The report decides which agent gets each fix:

- **Fix in docs** starts a Docs Agent session. The agent first checks that the problem is real and not already covered, and improves an existing page rather than adding a duplicate.
- **Draft in Marketing** starts a Marketing Agent session for content outside your docs, such as a blog post or a comparison.

For each path in detail, see [Turn a gap into a fix](/journeys/turn-a-gap-into-a-fix/) and [Write a post that answers a buyer question](/marketing-agent/buyer-question-post/).

## 4. Review and ship

Open the session from **View in Docs Agent** or **View in Marketing Agent** and read what the agent drafted.

- **Docs Agent:** review the change in the editor, then click **Raise PR**. Your team reviews and merges the pull request as usual. See [How EkLine opens pull requests](/agent/how-ekline-opens-pull-requests/).
- **Marketing Agent:** edit the draft, then click **Copy** or **Download** and publish it on your site yourself.

The assistants reach your pages only through the public web, by searching and fetching them, so a fix can change their answers only once it is live on your public site.

## 5. Run the collection again

Open the **Buyer questions** collection from the **Collections** tab in **Journeys** and click **Run collection**. Nothing runs it on a schedule.

The first run of a collection is its baseline, and each later report is compared with it. The headline says how many more or fewer questions you win than on the baseline date. A question whose outcome changed shows what it was, such as **Was a loss**. See [Track results over time](/journeys/track-results/).

## Check the result

- The fix's button on the report reads **View in Docs Agent** or **View in Marketing Agent**.
- Your team merged the pull request, or you published the post.
- In the new report, the **Passed** column for Acme's split payouts question shows the change since the baseline. If the question moved out of **You lose**, it shows **Was a loss**.

Answers vary from run to run even when nothing on your side changed: the assistants search the live web, and neither their answers nor the grading are fixed. Judge the change by the question's passed count across runs, not by a single answer.

## Next

- [Read a Buyer questions report](/journeys/read-a-report/)
- [Track results over time](/journeys/track-results/)
- [Turn a release into docs, release notes and a launch post](/get-started/recipes/release-to-launch/)