ADO Pilot runs a review automatically every time you push to an open PR. When you want a fresh review without a meaningful code change — for example, after fixing a transient error or tweaking review settings — you can push an empty commit, use the PR action menu, or trigger one from the Settings Hub.

{% callout type="info" %}
Not seeing any automatic reviews at all? If your organization connected without setting up service hooks yet, pull requests won't be reviewed on push until hooks are configured. Use the **PR action menu** or **Settings Hub** below to request reviews manually in the meantime — both work without service hooks.
{% /callout %}

## Automatic — review on push

The default setting `reviewOnPush: true` makes every push to an open PR trigger a review:

1. You push a commit (or open the PR for the first time).
2. Azure DevOps fires a `git.pullrequest.created` or `git.pullrequest.updated` service hook to ADO Pilot.
3. ADO Pilot queues a review against the latest commit.
4. The tracking comment on the PR walks through queued → in progress → complete.

You do not need to click anything. Push and wait.

## Manual — push an empty commit

When you want to re-run the review against an unchanged codebase — for example, after fixing a transient error or changing a review setting — push an empty commit:

```bash
git commit --allow-empty -m "trigger ADO Pilot review"
git push
```

This is the right approach when:

- The previous review errored (`PROVIDER_RATE_LIMITED`, transient model failure) and you want to retry without producing a substantive code change.
- You changed an exclusion pattern or a review setting and want to see the effect on this PR.

## Manual — PR action menu

{% figure src="/images/reviews-rerun-action-menu.png" alt="Clicking Run AI Review in the PR overflow menu to trigger a fresh review without pushing a commit." caption="Re-running a review via the PR action menu" /%}



When the ADO Pilot extension is installed, a **Run AI Review** entry appears in the three-dot overflow menu on any pull request (the same menu that hosts **Mark as draft** and **Abandon**).

Selecting it opens a confirmation dialog:

> Run an AI review of pull request **#123** now?

Click **Run review** to start the review. Click **Cancel** to dismiss without doing anything. After clicking **Run review**, the dialog settles on one of these outcomes:

- **Review started** — the review is now queued; inline comments and the status check post in a few minutes.
- **Review already in progress** — if you click again within the same dialog session, the duplicate is collapsed server-side and you are not charged again. A deliberate re-run (a fresh dialog, or a new push) starts a new review and consumes review credits.
- **Too many review requests** — you're triggering reviews too quickly for this PR. Wait a moment and try again; there's no **Try again** button here, since an immediate retry would just hit the same limit.
- **Review credits exhausted** — your org has used its review allowance for this period. A **Manage your plan at adopilot.dev** link appears for a billing owner to top up or change the plan.
- **Couldn't authorize the review** — ADO Pilot could not confirm you have access to this PR. Use **Try again** after verifying you are signed in with access to the PR.
- **Finish setting up ADO Pilot** — this organization hasn't completed onboarding yet, so there is no ADO Pilot account to review against. A **Sign up at adopilot.dev** link starts signup; no credit is used.
- **Your plan doesn't include this** — on-demand reviews via a service principal need a higher plan for this organization. A **Manage your plan at adopilot.dev** link appears so a billing owner can upgrade.
- **This organization is inactive** — AI reviews are turned off for this organization. A **Manage your plan at adopilot.dev** link lets you check subscription status and reactivate them.
- **Manual reviews are paused** — an administrator has temporarily disabled on-demand reviews. Automatic reviews still run on every push.
- **Something went wrong** — a transient problem. Use **Try again** in a moment.

Opening the dialog costs nothing. Only clicking **Run review** spends review credits (one per ≈500 billable diff lines, one-credit minimum).

## Manual — Settings Hub

Any organization member can also trigger a review from the Settings Hub: **Organization Settings → ADO Pilot**.

The hub shows a **Run a review** form when the org status is **Active**. Paste a full pull request URL (for example, `https://dev.azure.com/org/project/_git/repo/pullrequest/123`) and click **Run AI review now**. The form shows the same result messages as the PR action menu dialog above, plus a few URL-specific errors if the pasted URL is malformed, points at a repository you don't have access to, or its organization can't be determined.

The Settings Hub is useful when you want to kick off a review for any PR across your org without navigating to that PR first.

## Manual — retry from the dashboard

Every review has a detail page in the dashboard at `https://app.adopilot.dev/dashboard/reviews/{reviewId}`. Failed reviews show the error and a message directing you to retry the PR. There is no one-click re-queue button on the review detail page — use the PR action menu or push an empty commit to re-trigger.

You will use the dashboard review detail path when:

- You found the original review through a search or notification rather than from the PR itself.
- The PR has scrolled past your immediate workflow but you want to re-examine why the review failed.

Successful reviews (PASS, ADVISORY, FAIL) do not need a retry — there is nothing to re-queue. To run a fresh review on the same PR, push a new commit (or an empty one).

## Disabling automatic reviews

If you prefer to control timing yourself — for example, your team batches changes and only wants a review on the final commit — turn `reviewOnPush` off:

1. In the dashboard at `app.adopilot.dev`, go to **Repositories**.
2. Find the repo and toggle **Review on push** to **off**.
3. Save.

With `reviewOnPush` disabled:

- Pushes no longer trigger reviews. The webhook still arrives at ADO Pilot, but the review is skipped — no review record is created for a skipped push, and it doesn't count against your plan.
- To get a review, use a manual trigger instead: the **Run AI Review** PR action menu or the Settings Hub **Run a review** form. Both run on the manual lane and bypass `reviewOnPush`, so they review the PR even with automatic push reviews off.
- The dashboard review detail page still shows any reviews that did run before the setting changed — use the PR action menu or Settings Hub above to start a new one.

This setting is per repo, so you can mix automatic and manual repos in the same org.