Turn ranked AI visibility actions into Asana SEO tasks with Claude

Most teams start from a template. Asana lists SEO templates for strategy, roadmaps, audits, rank tracking and keyword research, and describes the SEO roadmap template as a way to plan audits, keywords, content and technical fixes, assign owners, set priorities and report KPIs. The structure is solved. A project, sections, owners and due dates are enough to run SEO work.

What the templates cannot supply is the content of the rows. Someone has to decide that a pricing page needs a refresh or that a comparison article is missing. That decision usually comes from a person's memory of a tool they checked last month, and it goes stale. This page keeps the Asana structure and replaces the guesswork with a queue of measured findings.

Teams lean on checklists because they work. An SEO consultant on Reddit wrote about the first item on theirs: "This alone fixes 80% of the technical issues I see on new sites. The other 20% is site-specific." The checklist covers the technical half. The Answer Radar cards cover the half a checklist cannot see, which is where answer engines name a competitor instead of you.

Each card is a short label plus the evidence behind it. In the live run, the top card reads that Umami holds the answer for one comparison question, with competitor_mentions of 3 and competitor_citations of 12, and the paired card says the brand is not represented for the same question, with samples_absent of 3. Both carry a confidence of 1 and an intent_score of 0.6.

That is enough to write a useful task. The title is the question. The description says which competitor an engine named, how many sources it cited and in how many samples the brand was absent. Linkeddit supplies the facts and stops there: its queue carries no written brief, so the customer's own assistant writes the task, and you edit it.

In the run, the competitor_won card and the content_gap card share the same gap_id and the same question. Two cards, one piece of work: a page that answers the comparison better. If Claude creates a task for each, the backlog counts one problem twice and two people might each build an article.

So the prompt asks Claude to group cards by question before drafting. The Asana check does the other half. Searching with search_tasks_preview first means an open task is updated with update_task instead of copied. A short backlog with one owner per question gets finished more often than a long one.

Claude proposes it. The Asana connector's listing names create_task_preview and create_project_preview as its creation tools, alongside update_task, so the connector is built around showing a draft before it becomes real. We did not run the Asana side, and what happens after a preview depends on your Asana workspace and your Claude settings.

Treat the preview as the review step. Check the title, the owner, the due date and the evidence, and look for a claim the Linkeddit data does not support. The prompt ends with a rule: no card moves to in_progress until you have approved its task, so Linkeddit and Asana never disagree about what is under way.

Say so. The live run returned two cards, and Linkeddit's own note says only content_gap and competitor_won are ever generated: the other five card types exist in the schema and have no evidence source yet. A short queue is not evidence that there is nothing else to do.

Keep that in the task notes. If the audit behind the queue finished partial, as the test project's recent runs did, put that on the Asana project description so nobody reads a small backlog as a clean bill of health. Linkeddit also records a dismissed card as a permanent suppression, so dismiss only what you have decided not to pursue.

Add them as a second source, once Search Console is connected to Linkeddit. gsc_queries_get lists queries ranked by impressions and has views for average position 11 to 20, top five positions clicked less than expected, and queries with no dedicated page. A query on page two with real impressions is a refresh task: one existing page, one query, one owner.

Use a separate section in the same Asana project so the two kinds of work stay readable, one headed AI answers and one headed Google queries. Ask Claude to label each task with its source, so that three months from now you can tell whether the answer engines or Google drove the work.