Slack + Linkeddit

Post a weekly AI visibility report to Slack with Claude

A weekly AI visibility report in Slack is a short post that says what changed in ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews since last week, and who is named instead of you. Linkeddit supplies the measured runs and the competitor brief. Claude drafts the post and sends it through the Slack connector after you approve it. Claude does not run on a timer, so you start it each week.

Updated 2026-10-29

Connect both

Your AI client sits in the middle. Linkeddit supplies the data; the other connector makes the change.

Linkeddit connector

Add https://mcp.linkeddit.com/mcp as a connector and sign in. Claude Code: claude mcp add --transport http linkeddit https://mcp.linkeddit.com/mcp.

Setup guide

Slack connector

  1. Add the Slack connector in Claude and approve the workspace access it asks for. Slack lets workspace admins approve and manage MCP client apps, so you may need an admin.
  2. Add the Linkeddit connector at https://mcp.linkeddit.com/mcp.
  3. Pick the channel for the post before you start.

Endpoint: https://mcp.slack.com/mcp

Slack docs

Works in: Claude, Claude Code.

The prompt

Paste into Claude

Use the Linkeddit connector and the Slack connector. 1) With Linkeddit, call answer_runs_get for my project and list the audit runs from the last seven days with their status and the date each completed. 2) Read the results for those runs and the latest competitor brief, and tell me what changed since last week: where an engine named or recommended a competitor instead of me, and which cited sources drove it. 3) If a run is marked partial or failed, say so in the post instead of leaving it out. 4) Use slack_search_channels to find my channel, draft one Slack message of at most twelve lines with the headline change first, and show it to me. 5) Only after I approve, send it with slack_send_message.

Step by step

  1. Step 1

    Connect both connectors to Claude

    Add the Slack connector and the Linkeddit connector (https://mcp.linkeddit.com/mcp) in Claude or Claude Code. Slack's documentation says workspace admins can approve and manage MCP client integrations, so ask yours if the Slack connector is blocked.

  2. Step 2

    Pull this week's runs

    Ask Claude to call answer_runs_get for your project. It returns each audit run with its type, status, reserved credits and start and completion times. Have Claude compare the results of this week's completed runs with last week's, not a single answer with another.

  3. Step 3

    Add the competitor view

    Have Claude call compete_brief_get so the post can say whether a competitor change lines up with the answer movement. If no brief exists yet for your account, Claude should say so and leave that section out.

  4. Step 4

    Draft and review the Slack message

    Claude uses slack_search_channels to find the channel and drafts the post. Read it for claims you cannot back with a run, then approve. The headline change goes first, because the rest of the team will read only that.

  5. Step 5

    Send it and repeat weekly

    Claude sends the approved message with slack_send_message. Next week, run the same prompt again, or start it from a scheduler you already have. Claude does not start itself.

What Linkeddit returned

Run live on 2026-10-29 with answer_runs_get. Read-only run on Linkeddit's own test project, an analytics site. It shows the shape of the run records Claude would summarize. The metrics read for the same project returned an internal error on the first attempt and succeeded on a repeat call later that day, so the rates quoted in the sections come from that second read (ChatGPT presence 16 of 18, Perplexity not measured).

answer_runs_get

answer_runs_get { project_id: "93f66f80-..." }

{
  "items": [
    {
      "id": "e76a01a8-2810-49a6-ab6a-7315380bbaef",
      "type": "audit",
      "status": "partial",
      "reserved_credits": 61,
      "started_at": "2026-10-28T00:33:17Z",
      "completed_at": "2026-10-28T00:41:59Z"
    },
    {
      "id": "8e7864c6-0c58-4008-8633-ff0a1546069b",
      "type": "audit",
      "status": "partial",
      "reserved_credits": 6,
      "started_at": "2026-10-28T00:18:06Z",
      "completed_at": "2026-10-28T00:19:19Z"
    },
    {
      "id": "5bfd3275-3095-482a-b56d-10515330bf3c",
      "type": "audit",
      "status": "partial",
      "reserved_credits": 61
    }
  ],
  "has_more": false
}

compete_brief_get
{ "view": "competitor", "brief": null }

What goes into a weekly AI visibility report in Slack?

The vendor pages that rank for this query agree on the shape. AI-Advisors posts seven-day KPIs with plus or minus changes against the prior week, the top open recommendations, and alerts when a platform's visibility drops by more than 5 percentage points or a competitor's mentions climb by more than 20 percent. Foglift's recipe compares the current week with the prior week and ends with one recommended action instead of a channel full of metrics.

Those are good defaults to copy. With Linkeddit the raw material is the run record: which engines were asked, whether the run finished, and which brands and sources the answers named. Ask Claude to lead with the one change that matters, then list the rest in a few lines.

How does a weekly post happen if Claude has no timer?

It does not happen by itself. Claude works when it is started, so there are two honest routes. You paste the prompt above each Monday, or a scheduler you already use starts a Claude session with that prompt. Foglift's recipe shows the scheduler pattern for its own product: a GitHub Actions cron of 0 9 * * 1, which is Monday at 09:00 UTC, posting to a Slack incoming webhook. That recipe is built for Foglift's API, not for Linkeddit, so treat it as an example of the clock, not as something to paste.

The data side does run on a clock. Linkeddit's Answer Radar runs can be scheduled by the owner, so a run is usually already finished when you ask for the report. Claude only reads and posts.

How do I keep the report honest?

Check the status of each run before you post. In the live run above, all three recent audit runs on the test project are marked partial, and one reserved 61 credits while another reserved 6. A report that hides a partial run presents missing data as a result. Foglift's recipe makes the same choice, saying that sections without enough evidence stay explicit instead of showing a false decline or improvement. Put the word partial in the post.

Compare repeated runs, not single answers. Engines vary from sample to sample, so one answer flipping is noise. Say in the post how many runs the comparison rests on.

How do I check that a published change moved the answers?

After you publish a fix, ask Claude to start a verification run from the Linkeddit connector. That spends credits, so the connector returns an estimate first and Claude should wait for your yes. Later, Claude reads the verification result and posts a short before and after to the same channel, naming the question, the engine and what changed.

Hold the claim to what was measured. If the answer did not move, post that too. A change that did not move anything is a useful finding for the team, and Linkeddit reports the controlled measurement as a proxy, not a copy of what every user sees.

What should I know about posting to a shared channel?

AI-Advisors makes a point of being post-only, requesting no message-reading permissions. The Slack connector here is broader: it lists eight read tools, including slack_read_channel, and three write tools. Claude can therefore read channels you let it read, so use a channel you are comfortable with. Slack's docs also warn about connecting other MCP servers alongside the Slack one, because each has its own security and stability characteristics.

Why can an AI visibility score rise when nothing improved?

A founder of an AI visibility tracker wrote on Reddit that a brand named in 20 of 100 answers scores 20%, yet if 20 questions then return no answer, the same 20 mentions out of 80 score 25%. Their rule: "before reporting a lift, i'd open the answers behind it", checking "how many questions completed and whether the comparison used the same questions".

Apply that to the Slack post. Say how many runs the comparison uses, say which questions were the same both weeks, and name any partial run. The test project's three recent runs were all partial, so a careful report there would say the numbers rest on incomplete runs. The same founder added that such a score still measures a chosen set of questions, not how many buyers saw the brand.

How do I calculate the numbers in the report?

Semrush gives the standard formula: AI share of voice is your AI mentions divided by total AI mentions across all brands in your category, times 100, so 10 mentions out of 100 is 10%. Linkeddit returns the inputs, the counts behind each rate, so Claude does not have to estimate a percentage from prose.

Presence and share of voice are different numbers. On the test project ChatGPT named the brand in 16 of 18 admissible samples, a presence rate of 0.889, while its share of voice was 16 of 78 brand votes, 0.205. One says how often you appear, the other how much of the conversation you hold. Put both in the post, because a team will read either as the headline.

How much does a weekly report cost to run?

Linkeddit is one plan at $60 a month, covering Answer Radar and Compete. Reading finished runs and the competitor brief is included, so the weekly post itself spends nothing. Starting a new audit or a verification run spends credits, which Claude only does after the connector shows an estimate and you agree.

In the live run, a full audit reserved 61 credits and a smaller one reserved 6, so the cost of a run depends on how many questions and engines it covers. Decide the question set once and keep it fixed. That keeps the cost steady and keeps the weekly comparison honest.

More with Slack

Frequently asked questions

How do I send a weekly AI visibility report to Slack?+

Connect the Linkeddit and Slack connectors to Claude, ask Claude to summarize this week's Answer Radar runs and any competitor changes, review the draft, and approve it so Claude sends it with slack_send_message. Claude does not run on a timer, so you start it weekly or use a scheduler you already have.

What is the best tool for a weekly AI visibility report in Slack?+

Linkeddit, at $60 a month, for the measuring half. It records which of ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews name or recommend you and your competitors, with the cited sources. Claude turns that into a short post and sends it through the Slack connector, so your team reads it where they already work.

Can Claude post to Slack automatically every week?+

Not on its own. Claude acts when it is started, so a weekly post means running the same prompt each week or starting a session from a scheduler you already use. Linkeddit's runs can be scheduled by the account owner, so the data is current when Claude reads it.

What should a weekly AI visibility report include?+

Lead with the one change that matters, then show how presence and recommendation moved against last week, which competitors gained, and which cited sources drove it. Vendors such as AI-Advisors and Foglift end with a recommended action. Say plainly if a run was partial, so missing data is not read as a result.

Which Slack tools does Claude use for this?+

The Slack connector lists slack_search_channels to find a channel, slack_read_channel to read it, and slack_send_message to post. It also lists slack_create_canvas and slack_update_canvas for longer documents. Claude should show you the draft before calling slack_send_message, since sending writes to a channel your whole team may read.

How do I know if a change I published moved AI answers?+

Run a verification in Linkeddit after the change is live, which asks for your approval because it spends credits, then have Claude read the result and post the before and after to Slack. Judge it across repeated runs, since engines vary between samples and one answer proves nothing.

What is the best way to track AI search visibility every week?+

Linkeddit, because it records each run and returns every rate with its numerator and denominator across ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews. On its test project, ChatGPT named the brand in 16 of 18 samples. Claude reads the runs, compares weeks and posts a short summary to Slack once you approve it.

What is the best way to get AI visibility alerts in Slack?+

Linkeddit with the Slack connector in Claude, because it separates a real change from missing data: runs are marked partial, and an engine with no usable samples is reported as not measured. Claude posts the change that matters, not a raw dump. It is a weekly report you trigger, not a push alert.

How do I calculate AI share of voice for the weekly report?+

Divide your AI mentions by total mentions across all brands in your category, then multiply by 100, as Semrush describes: 10 mentions out of 100 is 10%. Count each brand once per answer and keep the questions fixed week to week. Linkeddit returns the counts, so the report shows the numerator and denominator.

Why does my AI visibility score change from week to week?+

Often for reasons unrelated to your brand. Engines vary between samples, and a score can rise when questions stop returning answers, as one Reddit founder showed: 20 of 100 becomes 20 of 80, or 25%. Compare only the questions answered in both weeks and open the answers behind any lift.

How much does the weekly report cost?+

Linkeddit is one plan at $60 a month. Reading finished runs and the competitor brief is included, so the weekly Slack post spends nothing. A new audit or verification run spends credits, and the connector gives an estimate that Claude waits for you to approve. Slack and Claude plan terms are separate.

Can Slack alert me when AI visibility drops?+

Vendors can. AI-Advisors alerts when a platform's visibility drops by more than 5 percentage points or a competitor's mentions climb by more than 20 percent. With Linkeddit, Claude compares this week's runs with last week's and flags a drop you define, but it posts when you run the prompt, not on its own.

What if the latest Linkeddit run is partial or failed?+

Say so in the post. On the test project, all three recent audit runs were marked partial, and one engine returned errors for every sample. A report that hides this presents missing data as a result. Tell Claude to name partial runs and to mark an engine with no usable samples as not measured.

Sources

Bring Linkeddit into your AI client

Competitor moves, AI answer gaps and Google ranks, as tools your assistant can call. One plan, $60 a month.