How to get useful work done with your AI

Start with a connected account and active plan. Give your AI the website to manage. Ask for one useful change at a time and review changes before it saves them. Available actions depend on your plan, permissions and connected services.

Understand a result

Use Akkrev for https://example.com/. Read my latest report and the original
answers for one question where we were rarely recommended. Show the citations,
note missing samples and explain what we can and cannot conclude.

The AI reads results_get, then evidence_get with the returned report and question IDs. It can select an earlier report or inspect a partial report. Compare weeks only when Akkrev marks them comparable. API samples can differ from what a person sees in an AI app; a score change does not establish its cause.

What you get: The original sampled answer and its sources, with an explanation of the evidence and its limits.

Improve the tracked questions

Read my current tracked questions and allowance in Akkrev. Suggest one question
that matches what our guests actually ask. Show what you would add or remove
and wait for my confirmation.

After confirmation, questions_update adds or removes one question within your plan limit. Past reports retain their original question snapshots. Changing questions affects future scheduled samples and can make two weeks unsuitable for direct comparison.

Confirm accurate business facts

Read all my saved Akkrev facts first. I want to update our opening hours.
Ask me for the current hours and a public source, keep every unchanged fact
with its source, and show me the complete replacement set before saving.

facts_save replaces the entire saved set. Omitted facts are removed. Each nonempty fact needs its own public source URL and your confirmation. Keep temporary information current and use a review date when preparing a time-sensitive improvement. An AI answer is not proof that a venue fact is true.

Prepare a website improvement

Read my current Akkrev priorities and pick one supported task. Explain its
finding and completion check. Ask for any missing facts, then show the proposed
wording before preparing it. Keep publishing for my review in Akkrev.

The AI uses the current task ID and revision from workspace_get. It can start, snooze, resume or dismiss a task, prepare a reviewed change, or read a work brief. For a WordPress proposal it first prepares the wording, then uses an owned connection ID from connections_get and a matching page from website_pages. Grow and Scale support the included page proposals. Review and apply a proposal in Akkrev; preparation does not change WordPress.

For other website platforms, use your website editor or site manager. Your AI can explain the reviewed instructions. A standalone WordPress title or confirmed-facts proposal can use website_prepare when there is no matching tracked task.

What you get: A saved proposal or work brief, plus the relevant Akkrev review link. Your website is changed only through the applicable approval and implementation step.

Check a completed update

I have made the reviewed website change. Read the current Akkrev task again,
show me what will be recorded, and request its public check after I confirm.

improvement_check records your confirmed manual update or checks an update already recorded. Queued, awaiting verification and publicly verified are different states. Read workspace_get later for the outcome; avoid rapid polling. A public check confirms a supported page change, not indexing, ranking improvement or a booking.

What you get: A recorded update and its public-check status. A queued check remains pending until the returned result confirms it.

Review an article draft

Read one existing unpublished Akkrev article draft. Check it against my confirmed
facts, suggest clearer wording and show the full edit before saving a draft.
I will review publishing in Akkrev.

article_get reads a draft selected from the workspace. article_save edits an existing unpublished draft on Grow or Scale. This does not create additional drafts outside your allowance. Article publishing, provider delivery and automatic-publishing consent remain in the dashboard.

Review Google reports and listings

Read my connected Google report status and saved reports in Akkrev. Explain the
available data. If a matching property is already authorised, ask before selecting
or refreshing it. List any connection step I need to finish myself.

connections_get returns supported website connections and available Google properties/reports. google_refresh uses an already connected matching property and the normal refresh interval. You authorise Google and configure Analytics event mapping in Akkrev. No connected account means the AI returns the connection step; it cannot sign in on your behalf.

For listings, listing_review_save records the details actually checked against the correct public venue profile. It does not edit an external listing. Ask for a recorded review only after the check was made.

Follow the work

Akkrev continues its existing scheduled work after the conversation closes. Your AI can read job and task status when you return. Do not assume that the chat itself will stay active or send updates. Ask your AI to report saved, queued, applied and verified work separately. See the tool reference for exact inputs and permissions for approval boundaries.