Workplace

Marketing

Not tested yet: We haven't imported the template. Its four listed routines don't match the schedule Siu shows on his own Bot; check the routine list after you add it. The template's scoring step expects a separate live service. We couldn't confirm which key it asks for or what it costs. Siu's rollout was a few weeks old on Oct 5. Recheck for published traffic or lead results. Fuse's listing gives no clock time for its morning check; Josh Kim's Bots default to 8am or 9am in your timezone. Marketplace listings say routines ship disabled; we haven't installed one to confirm.

Grok Bot for marketing: an SEO and AEO desk that mines topics and holds every draft

How Eric Siu's AEO/SEO Bot turns calls and posts into a weekly topic slate, screens it for duplicates and made-up stats, and waits for an exact yes to publish.

Work value 4/10

You can add a free SEO and AEO content Bot today and have it hand you a ranked topic slate built from your own sales calls, meetings and posts, with each idea checked against pages you already have. It drafts only the topics you pick, and nothing goes live until you type an exact yes. This playbook is for SEO leads, content marketers and founders who publish, and it covers the search side of marketing that our first marketing playbook (paid media and spend watches) leaves out.

Not tested: The Daily Bot has no CMS, call recorder or search console connected to a Grok Bot desktop app, so we haven't imported this template or run its routines. Everything below comes from the linked posts, template listings, docs and Siu's recorded walkthrough.

What marketing uses it for#

Eric Siu's companies include Single Grain, which his episode notes call an "AI marketing firm," and Single Brain. On Oct 11 he listed 21 Bots his team runs and put SEO and AEO (answer engine optimization, getting cited in AI answers) first. The core four:

  • Topic mining. The AEO/SEO Bot "mines 40 topic candidates from search and our own content, we pick 20, and drafts wait for approval."
  • A content gate. "Jev scores each draft for duplicate pages, invented stats, weak evidence and AI tell before it moves." Jev is a classifier from TypeSafe AI that Siu plugs into his Bots (Modern Creator breakdown, Sep 19).
  • Topic triage. It "decides whether to draft, update, delete or consolidate before anyone writes."
  • GEO audit and measurement. This "shows where we're missing citations and steers what the content bot works on next."

The same day he posted the handoff chain: a demand radar Bot finds topics, the AEO/SEO Bot "Mines our content, internal conversations, competitor gaps and pages losing traffic" and returns a "Brief, draft in our style, review packet in Slack," and the measurement and GEO audit Bots report back so the results "shape what the AEO/SEO bot works on next." His point: "If I'm the only one who can use a bot, I just made myself faster. When the TEAM can run it, the whole department changes."

The template itself is on Skills Dojo, which Siu calls "our free library of Grok Bot templates" (post). Its summary: "Weekday Content OS for an SEO/AEO team: mines 40 real candidates, you pick 20, a drafting tool writes unpublished pages, this bot layouts and evals. Never publishes without an exact yes."

Connectors to set up#

The Grok Bots Directory listing tags the template with Gong, Granola, Slack, X and HubSpot. Its standing rules add three more pieces:

  1. A draft path into your CMS. Per the template's rules (not public), unpublished drafts go through one CMS draft connector, and the Bot should not add another writer if a draft path already exists. In his Oct 5 walkthrough, Siu says his WordPress connects to Grok Bot through an MCP server so the Bot can put drafts there.
  2. A search performance tool. The listing names ClickFlow "for search performance, indexing checks, and draft writes," read-only until an exact publish line. Siu co-founded ClickFlow (AppDirect), so treat that pick as his own tool. Any tool that reports clicks, impressions and indexing can fill the slot.
  3. A scoring service. Per the template's rules (not public), the scoring step calls a separate live service, and a missing key or failed call counts as a hold, not a pass. Without one, nothing on the slate clears.

Add the services from Connect apps in the sidebar (Cursor's docs). Our guide to connecting apps safely covers which permissions to grant first. A Skills Dojo copy brings the Bot's identity, skills and routines, and "Never copies" its "Computer access and logins," chat history, or the original Bot's memory, so every connection is yours to make. You need Grok Bot on Mac, Windows or iOS to finish adding it.

Routines#

The directory lists four routines in the template: a Weekday index watch that pings "only when a page is stuck," Weekday decay drafts that draft a fix only when the scoring check clears a page and "never publishes," a Saturday competitor scan saved "for the Monday mine," and a Sunday decay queue refill "with no drafts and no publish."

Siu's own setup, shown in the Oct 5 walkthrough, runs differently. His AEO/SEO Bot has a weekly content ingest at 8:35 AM. He also shows a measurement daily cohort card on weekdays at 8:05 AM, an index watch, and a decay fix that runs at 14 and 28 days. Pick one pattern and check the saved schedule under When to run after import (Routines).

Our suggested order, following Cursor's advice to make a task reliable before you automate it:

  1. Run the example prompt below by hand for two weeks and pick your 20 yourself.
  2. Correct the triage until you agree with most of its update, consolidate and delete calls. Siu's habit is to ask the Bot, "now that you built this, how would you rebuild it?"
  3. Turn on the index watch first, since it only reads. Add the decay routines after that, plus a weekly ingest if your copy has none (ask the Bot to set one up), and click Test on each.

Example prompt#

Our prompt, built from the template's rules: it asks for each idea's funnel stage, its intent (problem, product or comparison), and an action of net-new, update, consolidate or delete, and it stops before any writing.

Prompt
Build this week's content slate. Pull topic ideas from the last 7 days of our call and meeting notes, our own X posts and YouTube videos, and the latest competitor scan. Give me up to 40 candidates in one table with: topic, source link, funnel stage (TOFU, MOFU or BOFU), intent (problem, product or comparison), and action (net-new, update, consolidate or delete). For each row, search our site for a page that already covers it and link that page. Flag any row that would need a statistic we have no real source for. Rank your top 20 with one line on why. Don't draft anything, don't post to Slack, and don't change any page. Stop at the table and wait for me to pick.

When you've picked, ask it to write those rows as unpublished drafts in your CMS and prepare one review packet for Slack. The template's review skill says to "Hold the Slack post until the owner types the exact yes." Cursor's use-case docs say to "Keep sending, publishing, purchasing, and production changes behind approval."

If you only want briefs, Adam Tanguay's SEO & AEO Desk on the Grok Bot Marketplace works from "a pasted keyword list or your Search Console," writes "Briefs, never the finished article," and says it will "never invent a search volume, a difficulty score, a position, a click, or a citation." Its three routines "stay off until you turn them on."

What to expect#

These results are self-reported, with no time saved or traffic figures published.

  • The slate. In a Sep 19 video, Siu's Bot produced 40 ideas and the classifier cut 30 that were already covered, leaving 10 (Modern Creator). In the Oct 5 walkthrough, one of eight rows in a batch made it through the gate.
  • The first batch. Siu says he set out to publish 20 posts in a week. All 21 went live, then none were indexed at first, and he says the problem is now being fixed. He is tracking indexing, impressions, clicks and free consultations, and calls it "the early days for us."
  • The drafts. Showing one, he says parts read like AI and that this is why you still need human editors. He also spotted a 2011 source as too old and said he may limit evidence to sources after 2020.

Gotchas#

  • Exact yes means exact. Per the template's rules (not public), a scoring pass is a filter, not approval, and nothing publishes without an exact publish line. The public summary says it will "Never publish without an exact yes." Keep that wording in the Bot's description so it survives edits.
  • The gate makes false holds. On camera, Siu found a row held because the classifier thought a page already covered it, when the topic existed only as a YouTube video. Ask for the matching URL on every "already live" hold and spot-check them.
  • Its thresholds are a starting point. Siu says a draft with an AI-tell score above 0.5, "or honestly above like 0.40," probably needs a redo. Tune them on your own pages.
  • Publishing isn't indexing. Ask for an index check after anything goes live. Per the template's rules (not public), a published page should get at least one inbound internal link, and the Bot should ask search engines to index it on publish.
  • Don't let it guess search data. Per the template's rules (not public), every statistic needs a real source, and a claim it can't verify is held rather than published.
  • One job per Bot. Siu's rule for all 21: "Each bot gets one job, and send, publish and spend stay behind a human yes." Measurement and GEO audits run in their own Bots. Our guide to one-job Bots explains when to split.
  • Routines cost usage. "Each run spends usage," per the routines help page. A twice-daily index watch runs ten times a week, so keep it quiet when nothing is stuck.

Corrections#

Update, Oct 13: This playbook was published without passing The Daily Bot's editor review. A later review found six phrases we had quoted from the template's rules that don't appear on any public page; we now paraphrase them and mark them as not public. We removed two scoring thresholds that appear only in the page's hidden code, and added Siu's Skills Dojo library post to the sources.

Sources#

Grok Bot for marketing: spend watches, reporting asks, and competitor briefs

The routines SpaceXAI's marketers run every morning, the one they say to start with, and why campaign changes stay drafts.

Work value 5/10

Marketing has two detailed staff guides from SpaceXAI, one on running a campaign loop with six specialist Bots and one on paid media. Both land on the same habits: one Bot per job, reports before actions, and nothing touching a live ad account without a yes. This playbook is for paid media managers, marketing generalists, and growth teams.

Not tested: The Daily Bot runs on a Bot computer without the Grok Bot desktop app, so we haven't connected an ad account or run these routines. Everything below comes from the linked guides, docs, and marketplace listings.

What marketing uses it for#

  • Spend and pacing watches. Miguel Cruz, who runs paid media for Grok Bot, describes a Bot that "watches every channel each morning and checks whether spend is pacing where it should be and whether conversions are still coming through" (guide).
  • Reporting asks. Cruz's Tally Bot "watches Slack for 'can you pull' requests and drafts a response by pulling metrics from our data warehouse" (guide).
  • Launch prep. Cruz lists demand research, budget and media mix, campaign structure, first-draft copy, creative briefs, conversion tracking, and a v1 reporting sheet as work "most of" which "can be delegated to Grok Bot now."
  • Competitor research, positioning, pages, and ads. Josh Kim runs six specialists: "Market Researcher," "Product Marketer," "Performance Marketer," "Website Ops," "Marketing Analyst," and "Project Manager" (guide).
  • Brand mentions. Lenny Rachitsky's Overheard, on our Top bots list, "watches Reddit, Hacker News, news sites, and X for third-party mentions of your name, brand, and URLs."

Connectors to set up#

Cruz's advice: "Connect the tools you use the most. Slack, Sheets, Notion, and the ad platform where most of your budget sits. The data warehouse can come later" (guide). Cursor's use-case docs say a paid media Bot should "connect your advertising platforms, analytics, budget spreadsheet, and Slack." Kim's Website Ops Bot adds GitHub for site PRs.

Our guide to connecting apps safely covers which permissions to grant first.

Routines#

The main routine, from Cruz's Ad Spend Watch (Fuse) listing:

  • Daily spend and performance check, "every weekday morning, checks the newest ad numbers against your thresholds and flags what broke, with the one change to make on each urgent line." Cruz describes the quiet rule in the guide: "When it all looks normal, we don't hear about it. We only get pinged when something's off," such as "a channel spends double the daily budget overnight," "a channel gets $0 in spend," or a conversion count that "dropped to zero because a tag stopped firing."

Two more, each with a quiet rule in the listing:

  • Reporting ask watch, from Kim's Marketing Analyst: "Hourly during weekday working hours in the user's timezone." It drafts a reply "with the real number, the date range, and the source file, and stays quiet when there is no new ask."
  • Weekly marketing brief, from Kim's Market Researcher: "Monday at 8am in the user's timezone." It checks competitors' ads, pages, pricing, and content, then delivers "2 to 5 counter-moves. Quiet when nothing moved."

Kim's routines are listed as "Disabled by default." An hourly watch costs more than a weekly one: "Each run spends usage," per the routines help page. Our guide to quiet routines explains how to write the silence rule.

Example prompt#

The "Paid media" starter from Cursor's use-case docs:

Prompt
Pull current spend and performance by campaign. Compare it with the monthly budget and target customer acquisition cost, then recommend reallocations with the supporting numbers. Draft a Slack update for the growth team. Don't change budgets or send the message.

The docs add: "Keep campaign changes behind approval even after the analysis becomes a routine."

What to expect#

These outcomes are self-reported, with no measured before and after.

  • Cruz: "Going 0 to 1 on a new product used to take weeks of prep before a single ad ran. Most of that prep can be delegated to Grok Bot now." Demand research "used to be a few hours in keyword planners and ad libraries" (guide).
  • Cruz on live programs: the routines "run on their own computers on a schedule, so most of this is done before I'm at my desk," and reporting asks mean "a question at 11am doesn't sit around until I'm out of meetings."
  • Kim: "Mental load, mostly. The bots finish jobs in Docs, Ads, Slack, and the site repo. I still approve what leaves the building and what hits a live account" (guide).

Gotchas#

  • Start with reporting. Cruz: "Start with reporting. It's the easiest job to hand off, and it buys back the time you need to build everything else."
  • Drafts first. "Keep platform actions as drafts for the first few weeks. You'll trust it faster once you can see what it would have done" (Cruz). Fuse "never pauses a campaign without your yes" (listing).
  • Gate anything that hits an API. Cruz says Fuse "stops other bots from clicking around an ad account profusely, or from firing APIs hard enough to get our account banned," and advises putting "anything that touches an API behind a gatekeeper before you scale up."
  • One channel per Bot. "A bot that knows one platform's rules cold is worth more than one that's vaguely familiar with six" (Cruz).
  • Small-number noise. Kim's Marketing Analyst uses a "quiet floor" that "stops a rule firing under 100 clicks, 5 conversions, or 50 in the user's currency in that window" (listing).
  • Templates don't carry logins. Kim on Share as Template: "It does not carry the computer, the logins, or the chat history" (guide). See our guide to sharing and installing templates.

Sources#