This guide is published by Ryze AI (get-ryze.ai), which builds the Grok for marketers connector described here and sells an autonomous AI marketer for Google Ads, Meta Ads and SEO, so it has a commercial interest in the topic. It was first published by Ira Bodnar as an article on X on September 1, 2026, and collects what 1,000+ marketers in the Ryze community do with Grok Bot. The setup: add the Grok for marketers connector (x.ai/bot/dep-tU0gmIPgiqNsvS4N4) to Grok Bot, which exposes 36 tools across 50+ marketing data sources (seven SEO and GEO data providers, Google Ads, Meta Ads, Meta Ad Library, Search Console, Shopify, PostHog, Google Analytics, ChatGPT Ads, 16 image and video models, and CMS edits on Framer, Webflow, WordPress and Shopify), then run seven bots, one per lane: Google Ads (search terms, negatives, RSAs, quality score, pacing), Meta Ads (ad sets, budgets, the 50-signups rule, fatigued ads), Creative (30+ new ads a week, Ad Library scans, hooks, formats), SEO (Search Console, index coverage, new pages, site edits), GEO (whether ChatGPT, Gemini, Perplexity and AI Overviews cite you across 40 buyer queries), Tracking (GA4, PostHog, Shopify, conversion actions, platform-vs-analytics gaps over 10%), and Ops (onboarding, playbook, postmortems, the 5 a.m. meeting; never touches an ad account). Eighteen things to know: each bot has its own memory so one lane per bot; skills are short files read before every run with threshold, exceptions and proof; the search-terms skill is 30 days, $20 spend, 0 conversions, campaign-level negatives, never brand terms, never a term that converted in 90 days; proof is the list re-pulled, the URL inspected, the ad status checked, never an API saying OK; automations run on three clocks, every 30 minutes for money on fire, daily for slow leaks, weekly for the full audit; one shared Notion board logs every change with what, when and proof; bots change on their own only small reversible things (a $20/0-conversion negative, pausing a 72-hour-old Meta ad at 3x target CPA with zero signups, a 20% budget raise on a winner once a day, a sitemap resubmit, a title fix on a page with no rankings); anything over 20% on budgets or bids, new campaigns, pausing anything that spent over $1,000 last week, new ad copy, conversion tracking and edits on money pages wait for approval. The page includes the full paste-in prompts for all seven bots and the community's 3 a.m. prompt. Updated October 2026. Updated October 19, 2026.
|
Ira Bodnar··Updated ·14 min read

7 Grok Bots for Marketing

18 things to know and 7 paste-in prompts for running ads, SEO and GEO on Grok Bot. Collected from 1,000+ marketers in our community. One bot per lane, three clocks, a shared board, and a short list of what the bots change on their own.

Add Grok for marketers to Grok Bot Jump to the 7 prompts

Built by our community of 2,000 marketers

Free skills and prompts for paid ads and SEO

Templates for Grok, Claude, ChatGPT and Perplexity.

Clients we work with

State Farm
Luca Faloni
Pepperfry
Slim Chickens
Superpower
Jenni AI
Tetra
Speedy
HG
Motif Digital
Grok Bot for Marketing cover: the Grok Bot mark next to the words Grok Bot for Marketing on an orange field

Start here

Add and connect the Grok bot for marketing

One click adds it to your Grok. Then connect Google Ads, Meta, Search Console, GA4 or Shopify, only the ones each bot needs.

Add to Grok →

Most of us use Grok Bot to run marketing work end to end:

  • Weekly SEO and GEO check, audits and fixes pushed straight to the GitHub repo or Shopify.
  • Weekly Meta check, dozens of new creatives generated and uploaded paused.
  • Google Ads search terms cleaned, negatives added, tracking verified.
  • Nightly check of what changed: rankings, ChatGPT citations, pages that dropped out of Google’s index.

The short version

→

Connect — one connector, 50+ marketing data sources, added to Grok Bot in a click

→

Seven bots — Google Ads, Meta Ads, Creative, SEO, GEO, Tracking, and Ops to run the team

→

Three clocks — every 30 minutes for money on fire, daily for slow leaks, weekly for the full audit

→

The line — small reversible changes run alone; anything over 20%, new campaigns, tracking and money pages wait for you

→

The loop — every change runs against a control; winners become rules in the skill file, losers are rolled back with a note

How to connect

  1. 1. Add Grok for marketers to your Grok Bot: https://x.ai/bot/dep-tU0gmIPgiqNsvS4N4. It gets access to all your marketing data.
  2. 2. You get the tools below to connect 50+ marketing data sources.
The Grok for SEO, GEO, paid ads and Shopify bot listing on x.ai: a senior marketer for Google Ads, Meta, SEO and Shopify, with the Add to Grok Bot button and the note that the bot was created by a third-party user and may act on your behalf
The connector listing. One click adds it to your Grok Bot.
Integrations screen behind the connector: Google Ads, Meta Ads, TikTok Ads, Reddit Ads and OpenAI Ads connected, with LinkedIn Ads, Microsoft Ads, Snapchat Ads, Pinterest Ads, Thrad, Criteo and StackAdapt available to connect
The advertising sources behind the connector. Search Console, GA4, Shopify, PostHog, seven SEO and GEO data providers and 16 image and video models sit on the other tabs.

Thirty-six tools in total: seven providers of SEO and GEO data (some are free), Google Ads, Meta Ads, the Meta Ad Library, Search Console, Shopify, PostHog, Google Analytics, ChatGPT Ads, 16 models for video and image generation, and the CMS connectors that make SEO edits on Framer, Webflow, WordPress or Shopify. The same connector runs from Claude and ChatGPT; the Grok setup is covered here, the Claude one in how to connect Claude to Google and Meta Ads.

The seven bots

3. One bot per domain. Six do the work, one runs the team.

Grok Bot chat list named Marketing with seven bots: Kit for Meta Ads (2 ad sets past 50 signups, +20% applied), Nora for Google Ads (180 negatives in, list re-pulled 180/180), Ava for Creative (batch of 30 uploaded paused), Otto for SEO (crawled-not-indexed up vs yesterday), Vera for GEO (40 buyer queries re-run, cited on 23), Theo for Analytics (Google Ads 42 conversions vs GA4 38, tag fires twice on mobile) and Marta for Ops (5am 1:1s done, playbook v12 shared)
Seven bots, one lane each. The previews are last night's work.
#BotOwns
4Google Ads Grok Botsearch terms, negatives, RSAs, quality score, pacing
5Meta Ads Grok Botad sets, budgets, the 50-signups rule, fatigued ads out
6Creative Grok Bot30+ new ads a week, Ad Library scans, hooks, formats, previews
7SEO Grok BotSearch Console, index coverage, new pages, the edits on the site
8GEO Grok Botwhether ChatGPT, Gemini and AI Overviews cite you
9Tracking Grok BotGA4, PostHog, Shopify, conversion actions
10Ops Grok Botonboarding, playbook, postmortems, the 5 a.m. meeting. Never touches an ad account

Connecting each platform is a one-time step: Google Ads and Meta Ads each take a couple of minutes. The Ops bot is the one people skip and then regret. It reads the shared board every morning, runs a 1:1 with each bot, repeats the playbook and posts the five-line summary. It never touches an ad account. Platform-by-platform detail for the two ads bots is in how to manage Google Ads from Grok and how to manage Meta Ads from Grok.

Things to know for Grok Bot maxxing

11. Memory. Each bot has its own, so one lane per bot. Put your unit economics in it once. US Google Ads at $500–700 per customer fails for SaaS with LTV under $1.5k.

12. Skills. A short file the bot reads before every run. Threshold, exceptions, proof. Like an onboarding doc. The same idea as Grok skills for Google Ads and Grok skills for Meta Ads, written for a bot instead of a chat; the format itself is explained in the Claude marketing skills guide.

13. A skill, example. /search-terms, the whole skill: window 30 days, $20 spend, 0 conversions, campaign-level. Never brand terms, never anything that converted in 90 days. Skip if the tracking bot has an open tracking-gap row. Proof = negative list re-pulled today, 7-day search terms next Monday.

Nora, the Google Ads Grok Bot, reporting a weekly search-terms clean marked Done: task google-ads/us-search-negatives, 180 negatives added, +0 conversions, minus $4,200, with View changes and Open in Google Ads buttons, followed by the message that the negative list was re-pulled 180 of 180 in
A skill run, finished. The proof is the re-pulled list, not the API response.

14. Proof. An API saying OK is not proof. Proof is the list re-pulled, the URL inspected, the ad status checked.

15. Automations. The scheduled runs. Three clocks, because ad data lags: every 30 minutes for money on fire, daily for slow leaks, weekly for the full audit.

16. Shared board. Keep one shared table (Notion works) where every bot logs each change it makes: what, when, proof. You check one table instead of reading seven chats.

The shared Notion board, Marketing Work Streams, on two phones: Nora's task list (search terms weekly, brand budget +30% needs approval, RSA #2 in 4 ad groups, QS under 5 landing page mismatch, sitemap resubmit, Meta batch of 30, alternatives page draft) and the table view with Owner, Stage and Proof columns showing Working, Watching 3/3, Watching 2/2 and Ready for review rows with proof links
One table for seven bots. Owner, stage, proof. The Ops bot reads it at 5 a.m.

17. Auto changes. What the bots change on their own: add a negative keyword (a search term that spent $20 with zero conversions in 30 days), pause a Meta ad that is 3 days old and spent 3x your target cost per signup with zero signups, raise a winning ad set’s budget by 20%, resubmit a sitemap, fix the title on a page that has no rankings. Small, and undone in a minute if wrong.

18. Needs approval. What waits for you: any budget or bid change over 20%, any new campaign, pausing anything that spent over $1,000 last week, new ad copy, anything touching conversion tracking, any edit on a page that makes money.

Nora's Google Ads report in Grok Bot: negatives landed, campaign negative list re-pulled 180 of 180 in, Monday check for zero impressions in the 7-day report, brand CPA up 11% week over week under the 15% flag, with three Google Ads screenshots of the 180 negative keywords attached as proof, terms like dental assistant jobs, paralegal salary, free crm, jobs near me
Applied is not done. Done is the screenshot of the account after the change.
Bots do aloneWaits for you
Add a negative keyword: $20 spent, 0 conversions, 30 daysAny budget or bid change over 20%
Pause a Meta ad over 72 hours old at 3x target CPA with 0 signupsAny new campaign
Raise a winning ad set budget by 20%, once a dayPausing anything that spent over $1,000 last week
Resubmit a sitemapNew ad copy
Fix the title on a page with no rankingsAnything touching conversion tracking
Generate and upload creatives, pausedAny edit on a page that makes money

Always on and self-improving: the 4 loops

Two things separate a Grok Bot from a chat. It keeps working when you close the app, and it gets better at your account every week. Neither is magic. Both come from the pieces above: the clocks, the skill files, the shared board and the proof bar.

Four bots, one loop each. Step 08 is where the account teaches the bot.

Always on

The bots run on their clocks, not on your attention. The 30-minute clock catches money on fire, the daily clock catches slow leaks and the weekly clock runs the full audit. Memory keeps your unit economics between runs. The Ops bot opens every morning with the board. Nothing waits for you to start a chat, and anything outside the auto rules waits for you as a row.

Self-improving

Grok does not retrain itself on your account. What improves is the bot’s skill file. Every change runs against a control. After a fixed window the bot compares the result with the control. A winner is kept and written into the skill as a rule, with the date and the proof. A loser is rolled back, and the skill gets a line on what not to repeat. The next run reads the updated skill. That is step 08 in every loop: keep, roll back, learn.

#Google Ads GrokMeta GrokSEO GrokGEO Grok
01Model unit economics
Margin, LTV and break-even ROAS		Map angles and awareness
Offer, personas, target CPA		Map revenue pages
Which URLs and intents make money		Write the brand brief
What AI should say about you
02Audit tracking and structure
Enhanced conversions, consent mode		Fix signal quality
CAPI, event match quality, dedupe		Crawl and parse logs
Crawl budget, indexation, rendering		Build a prompt set
200+ buyer prompts by persona
03Mine search term n-grams
Cluster queries by CPA, not keyword		Tear down competitor ads
Hooks, formats, spend velocity		Cluster search demand
By intent and SERP overlap		Benchmark share of voice
Mentions and citations, 4 models
04Sculpt negatives and match
Shared lists, exact for proven winners		Build a creative matrix
Angle × hook × format		Gap-analyze competitors
Topics, links and SERP features		Trace citation sources
Pages and threads models trust
05Test RSA assets
Asset-level wins, not ad-level guesses		Test in a sandbox
ABO tests, graduate to Advantage+		Pay down technical debt
CWV, canonicals, schema, links		Fix wrong answers at source
Brand facts page, schema, llms.txt
06Bid to margin, not revenue
tROAS with margin labels in the feed		Kill on leading metrics
Hook rate, hold rate, CTR		Publish and refresh clusters
Hub pages, decay fixes, entities		Publish citable pages
Answer-first pages on your own site
07Run holdout experiments
Campaign splits, not before/after		Scale without resets
Budget +20% a day or duplicate		Measure against control
Search Console deltas vs untouched pages		Re-run the prompt set
Share of voice vs last week
08Keep, roll back, learn
Winners become rules for the next loop		Keep, roll back, learn
Winners become rules for the next loop		Keep, roll back, learn
What Google rewards becomes a rule		Keep, roll back, learn
What gets cited becomes a rule

How a result becomes a rule

  1. 1. Pick the control before the change. A campaign split for Google Ads, a sandbox ad set for Meta, untouched pages for SEO, last week’s share of voice for GEO. Before/after is not a control.
  2. 2. Log it on the board. The row holds the change, the control and the date the bot will check it.
  3. 3. Wait the window in the skill file. Short for Meta creative, longer for search terms, longest for SEO, where a page needs weeks before the data means anything.
  4. 4. Keep or roll back. Keep adds a rule to the skill. Roll back undoes the change and adds the reason, so the bot does not try it again next week.
  5. 5. Share it. The Ops bot reads the updated skills at 5 a.m. and repeats any new rule in the 1:1 with the bots it affects.

Paste this into each bot, under its lane prompt:

Prompt
After every change, open a row on the board with the control you will compare against and the date you will check it. On that date, compare the change with the control, not with last week. If it won, keep it and add it to your skill file as a rule, with the date and the proof. If it lost, roll it back and add one line to your skill file saying what not to repeat and why. Start every run by reading your skill file.

Free SEO/GEO audit

  • Surface technical SEO issues
  • Spot quick wins
  • Find duplicate and thin pages

Free · no credit card · instant

Prompts (paste these in)

One per bot. Connect only the tools that bot needs. Copy, paste, rename. All of them run on the same connector: https://x.ai/bot/dep-tU0gmIPgiqNsvS4N4.

Marta, the Ops Grok Bot, onboarding Otto the SEO bot in a group chat at 5:02 AM: vibe is short messages, one answer once, negatives auto, budgets over 20% queued, no exceptions; the playbook changed yesterday, skim before the 6am run, with the Notion playbook link; then asks anything stuck
The 5 a.m. 1:1. Ops repeats the playbook, asks what is stuck, moves on.

Google Ads Grok Bot

Prompt
Name this bot `Google Ads`. You are the Google Ads bot. You keep our Google Ads accounts clean and paced. You never launch spend. First, make sure Google Ads, GA4 and Google Sheets are connected. If not, prompt me to connect them before doing anything. Every 30 minutes: - flag any campaign over 120% of daily budget by noon, or under 50% by 6 p.m. - flag disapproved and "limited by policy" ads - flag any campaign with over $200 spend in the last 4 hours and zero conversions - check every ad's final URL returns 200 and still has the GA4 tag Every day at 7 a.m.: - CPA by campaign, last 8 weeks, week over week. Flag anything up more than 15%. - Google Ads conversions vs GA4, last 7 days. A gap over 10% that opened this week means a broken tag: stop optimizing that account and tell me. Every Monday at 6 a.m.: - search terms, last 30 days: any term with over $20 spend and zero conversions becomes a campaign-level negative. Never brand terms. Never a term that converted in the last 90 days. - top 10 keywords by spend with quality score under 5, name the landing-page mismatch. You may do alone: add negatives, resubmit the sitemap. Everything else (budgets, bids over 20%, new campaigns, new RSAs, conversion actions) goes to the board as "Needs approval". Proof: after every change, pull the report again and attach it. An API saying OK is not proof. Give me back: campaign, what changed, before, after, proof link.

Meta Ads Grok Bot

Prompt
Name this bot `Meta Ads`. You are the Meta Ads bot. You manage ad sets and budgets. You never launch a new campaign. First, make sure Meta Ads and Google Sheets are connected. Every day at 7 a.m.: - any ad set past 50 signups, under target CPA for 3 days and hitting its budget cap gets +20%. Once a day, never more. - nothing under 50 signups gets judged. Conversion counts under 72 hours old are provisional: check signups by UTM in PostHog before pausing anything. - pause any ad over 72 hours old that spent 3x target CPA with zero signups. - flag any ad with frequency up and CTR down 3 days in a row, and ask the creative bot for a replacement. You may do alone: the +20% bump, the 72-hour pause. Everything else goes to the board as "Needs approval". Proof: re-pull the ad set after the change and attach the screenshot. Give me back: ad set, action, spend, signups, CPA before and after.

SEO Grok Bot

Prompt
Name this bot `SEO`. You are the SEO bot. You keep pages indexed and ranking. You edit the site directly through the CMS connector. First, make sure Search Console, GA4 and the site (Framer, Webflow, WordPress or Shopify) are connected. Every night at 3 a.m.: - Search Console "crawled, currently not indexed" against yesterday. List the new URLs. If a money page is in the list, inspect the URL and request indexing. - every page published in the last 24 hours: does it look mass-produced (same structure as other pages, nothing new said)? Flag it. The August 2026 spam update removes sites like that from search and AI Overviews the same day. - rankings for the money pages, diffed against last night. Every Monday: - resubmit the sitemap after any title changes. - pages losing clicks 3 weeks in a row: propose a rewrite, queue for approval. You may do alone: sitemap resubmits, a title fix on a page with no rankings. Any edit on a page that makes money goes to the board as "Needs approval". Proof: URL inspected in Search Console after the change, screenshot attached. Give me back: URL, what changed, index status, clicks before and after.

GEO Grok Bot

Prompt
Name this bot `GEO`. You are the GEO bot. Your one number is how many of our 40 buyer queries cite us in ChatGPT, Gemini, Perplexity and Google AI Overviews. First, make sure the GEO data providers and Search Console are connected. The list of 40 buyer queries lives in the `queries` tab of our sheet. Every night: - re-run all 40 queries on every engine. Record which page is cited for each. Diff against last night and tell me what we gained and lost. Every Monday: - for every query where we are not cited: is our page indexed in Bing? No Bing index, no ChatGPT citation. Fix that first. - does our page title mirror the query wording? Title-mirroring pages get cited 4 of 4 times, neutral titles 1 of 6. Propose the new title. - no page at all? Draft a "best N tools 2026" or "[competitor] alternatives" page with the query as the title. Queue for approval. You may do alone: nothing on the site. Every title or page change goes to the board as "Needs approval". Give me back: query, engine, cited page, our status, proposed fix.

Creative Grok Bot

Prompt
Name this bot `Creative`. You are the creative bot. The bar is 30+ new ads a week. You can't guess the winner, you can only out-test everyone. First, make sure Meta Ad Library, Meta Ads and the image and video models are connected. Every Monday: - pull the active ads in our category from the Ad Library. Keep the hook-and-format combos that win most often for our product type. - generate 30 statics and 10 videos across the models. Every one rendered in 4:5 and 9:16. A headline cut off in 4:5 is rejected, not uploaded. - upload the batch paused. A human turns ads on. Every night: - for every ad the Meta bot flagged (frequency up, CTR down 3 days in a row), queue a replacement from the batch, tagged by hook and format. - every new ad from our 10 named competitors in the last 24 hours, tagged by hook and format. You may do alone: generate and upload paused. Never turn an ad on. Give me back: ad, hook, format, source, preview links.

Tracking Grok Bot

Prompt
Name this bot `Tracking`. You are the tracking bot. No other bot optimizes toward a number you haven't checked. First, make sure Google Ads, Meta Ads, GA4, PostHog and Shopify are connected. Every day at 7 a.m.: - Google Ads conversions vs GA4 conversions, last 7 days, by campaign. - Meta conversions vs PostHog signups (or Shopify orders), last 7 days, by campaign. - a gap over 10% that opened this week means a broken tag: open a "tracking gap" row on the board and tell the Google Ads and Meta bots to stop optimizing that account until it's closed. Every week: - every landing page that spent over $20: the tag fires exactly once on desktop and mobile. You may do alone: nothing in the accounts. You report. Give me back: source, platform number, our number, gap, which page or tag.

Ops Grok Bot

Prompt
Name this bot `Ops`. You are the ops bot. You run the team. You never touch an ad account. Every day at 5 a.m.: - a 1:1 with every bot: read its rows on the board, ask what's stuck, repeat the playbook (row first; negatives auto; budgets over 20% queued, once a day; done = a human saw the screenshot). - post a 5-line summary for me: what changed yesterday, what needs approval, what's stuck. When a bot makes a mistake: - find out why from its reasoning, write the rule that prevents it, add it to the playbook, tell every bot. When I ask for a new bot: - create it, share the playbook, and have the tracking bot check Google Ads vs GA4 before the new bot optimizes toward anything.
Theo, the Tracking Grok Bot, onboarding a new bot on the board and playbook: row first, every change becomes a Notion row at Stage Working before touching the account, never edit another owner's campaign; auto rules are negatives at $20, 0 conversions, 30 days, pausing ads over 72h at 3x CPA, sitemap resubmits; queued are budgets or bids over 20%, new campaigns, RSAs, conversion actions, money pages; proof bar is re-pull the report, screenshot, attach; tracking first, Google Ads vs GA4 before optimizing
The playbook, as one bot explains it to a new one. Row first, auto rules, proof bar, tracking first.

The 3 a.m. prompt

The community’s favorite nightly prompt:

Prompt
It's 3 a.m. and you have until 8. Find the one thing in this account I'd be embarrassed a client saw. Fix it if it's inside the auto rules, otherwise leave me the row.

The connector: “Grok for SEO, GEO, paid ads and Shopify” (https://x.ai/bot/dep-tU0gmIPgiqNsvS4N4). It reads live data and makes the changes. MCPs on GitHub work too. Set your seven up and tell us what breaks.

If you would rather not run seven bots, the same rules, clocks and approval limits are what Ryze AI runs as one system, with the audit trail on every change. This page first appeared as an article on X on September 1, 2026.

If you want the documentation behind this, the Ryze help center has pages on what Ryze AI is and AI ad management across Google, Meta, TikTok and LinkedIn.

Frequently asked questions

Do Grok bots improve themselves?
Not by retraining the model. They improve by updating their own skill files. Each change is measured against a control after a fixed window. Winners are written into the skill as rules with the date and the proof, and losers are rolled back with a note on what not to repeat. The next run starts from the updated skill.
Do Grok bots keep running when I close Grok?
Yes, once you set the scheduled automations. They run on three clocks, every 30 minutes, daily and weekly, and log every change on the shared board. Anything outside the auto rules waits on the board for your approval instead of going live.
What is a Grok Bot for marketing?
A Grok Bot is a persistent bot inside Grok with its own memory, connected tools and scheduled runs. A marketing Grok Bot is one set up for a single lane, such as Google Ads or SEO, with the connector that reads live account data, a short skill file that states its thresholds, and automations that run it every 30 minutes, daily or weekly.
How do I connect Grok Bot to Google Ads, Meta Ads and Search Console?
Add the Grok for marketers connector at x.ai/bot/dep-tU0gmIPgiqNsvS4N4 to your Grok Bot. It exposes 50+ marketing data sources, including Google Ads, Meta Ads, the Meta Ad Library, Search Console, GA4, PostHog, Shopify and ChatGPT Ads, plus image and video models. Connect only the tools each bot needs. MCP servers from GitHub work too.
Show 11 more questions
Why seven bots instead of one Grok Bot for everything?
Each bot has its own memory, so one lane per bot keeps the context clean: unit economics, thresholds and exceptions for that platform only. Six bots do the work (Google Ads, Meta Ads, Creative, SEO, GEO, Tracking) and one Ops bot runs the team, reads the shared board and never touches an ad account.
What can the Grok bots change on their own?
Small, reversible things: add a negative keyword for a term that spent $20 with zero conversions in 30 days, pause a Meta ad over 72 hours old that spent 3x target CPA with zero signups, raise a winning ad set budget by 20% once a day, resubmit a sitemap, fix a title on a page with no rankings. Anything else is queued for approval.
What needs human approval?
Any budget or bid change over 20%, any new campaign, pausing anything that spent over $1,000 last week, new ad copy, anything touching conversion tracking, and any edit on a page that makes money. The bot writes the row on the shared board as Needs approval and waits.
What is a skill in Grok Bot?
A short file the bot reads before every run: the threshold, the exceptions and what counts as proof. The search-terms skill, in full, is a 30-day window, $20 spend, zero conversions, campaign-level negatives, never brand terms, never a term that converted in 90 days, skip if the tracking bot has an open tracking-gap row, proof is the negative list re-pulled today.
What counts as proof that a bot did the job?
An API saying OK is not proof. Proof is the list re-pulled after the change, the URL inspected in Search Console, the ad status checked in the account, with a screenshot attached to the row on the board. Applied is not done; done means a human saw the screenshot.
How often should the marketing bots run?
Three clocks, because ad data lags. Every 30 minutes for money on fire: overspend, disapprovals, spend with zero conversions, broken final URLs. Daily at 7 a.m. for slow leaks: CPA week over week, platform conversions against GA4. Weekly for the full audit: search terms, quality score, sitemap, pages losing clicks three weeks in a row.
What is the shared board and why use Notion?
One table, Notion works, where every bot logs each change: what, when, owner, stage, proof link. You check one table instead of reading seven chats. The Ops bot reads the rows every morning at 5 a.m., asks each bot what is stuck, and posts a five-line summary.
Which Grok Bot prompt should I set up first?
Tracking first, then Google Ads. The tracking bot compares Google Ads and Meta conversions against GA4, PostHog or Shopify every morning, and no other bot optimizes toward a number it has not checked. A gap over 10% that opened this week means a broken tag, and the other bots stop optimizing that account until it is closed.
Can Grok Bot generate ad creatives?
Yes. The Creative bot pulls active ads in your category from the Meta Ad Library every Monday, keeps the hook-and-format combinations that win most for your product type, generates 30 statics and 10 videos across the connected image and video models in 4:5 and 9:16, and uploads the batch paused. A human turns ads on.
What does the GEO bot do?
It re-runs your 40 buyer queries every night on ChatGPT, Gemini, Perplexity and Google AI Overviews, records which page is cited for each and diffs against last night. On Mondays it checks Bing indexing for every uncited query, proposes title changes that mirror the query wording, and drafts best-N-tools or competitor-alternatives pages, queued for approval.
What is the 3 a.m. prompt?
The community's favorite nightly prompt: it is 3 a.m. and you have until 8, find the one thing in this account I would be embarrassed a client saw, fix it if it is inside the auto rules, otherwise leave me the row. It runs on the same connector and respects the same approval limits as the seven bots.

Related guides

How to manage Google Ads from Grok

The weekly loop, nine prompts, approvals and scheduling.

How to manage Meta Ads from Grok

Ad sets, budgets, the 50-signups rule, fatigued ads out.

Grok skills for Google Ads

The skill files the Google Ads bot reads before every run.

How to connect Grok to Google Ads

Step by step, with what the connector can and cannot touch.

Let AI run your ads

Google + Meta · audits 24/7

Start free trial
Live results across
2,000+ clients

Paid Ads

Avg. client
ROAS
0x
Revenue
driven
$0M

SEO

Organic
visits driven
0M
Keywords
on page 1
48k+

Websites

Conversion
rate lift
+0%
Time
on site
+0%
Last updated: Oct 18, 2026
All systems ok
PricingCompareConnect ClaudeDocsAbout
AI SEO toolAI visibility toolGEO toolSecurityChangelogAffiliatesAI InfoPrivacy PolicyTerms of ServiceRefund PolicySMS Opt-Inhello@get-ryze.ai