GrokBot as an SEO Agent

Source

Recirc that sent us here: Sarvesh Shrivastava (@bloggersarvesh) — x.com/bloggersarvesh/status/2093694576242696679 (29 Aug 2026). “open Grok Bot / paste these 20 prompts / dominate Google like a $10k/month SEO agency / make your first million dollars.” ~2.7k bookmarks / ~207k views at capture. 18s hook video. Quotes the 19 Aug X article Top 20 Grok Bot Prompts For SEO ( status/2090071557590900974 — ~3.8k bookmarks / ~1.25M views).

Same author, useful companion (not the 20 prompts): “STOP prompting Grok Bot like it’s 2024 ChatGPT” — business brain + COMPETITORS.md + buyer-intent filter before you paste anything. Earlier 90-day playbook: Grok Bot + SEO, 90 days . The consult CTA is Alventra Marketing. Steal the operating system; skip the funnel.

Honest replies on the article worth keeping: @AgileBrad — “a 2018 local-SEO system with 2026 Grok-Bot packaging.” @GregoryMcKeown on the recirc: “hard pass.” Both are useful filters. The prompts are still a decent punch list if you run them as a named bot, not a mega-chat.

Different artifact, don’t mix: Daniel Ch — “I told Grok Bot to do our SEO” ($27k MRR SaaS loop via Conqueror MCP + GSC, stop-before-publish). That’s closer to RankingSolution than plumber map-pack.

One-sentence TL;DR

One named SEO bot under the Chief, business context loaded once, live browser reads of GBP + site + GSC, DataForSEO instead of logging Ahrefs into the shared VM, human still hits publish — then schedule the monthly report, don’t paste twenty prompts into one Wednesday.

What this actually is

Fourteen years of local SEO for home services (plumbers, HVAC, lawyers, cleaning) packaged as Grok Bot computer-use prompts. Four parts:

1. GBP (1–8) — categories, attributes, reviews, posts, services, description, photos.
2. Website (9–13) — keyword gap, money pages, service+city pages, GSC page-2 sprint, review-language copy.
3. Authority (14–16) — competitor backlinks, NAP citations, buyer-journey keyword map.
4. Content + tracking (17–20) — content gap, entity/schema, GBP posting forensics, monthly report.

The move that makes it a Grok Bot job instead of a ChatGPT essay: almost every prompt starts with Open Chrome and go read the live listing / the live SERP / the live GSC export. That’s the product. The million-dollar line is marketing.

His own hedge, from a later recirc: SEO can’t be fully automated. The bot replaces the 40 hours of busywork that stops a local shop from doing SEO at all. Strategy stays yours. Execution gets faster. Eric Siu said the same thing from the other side: last mile is human, GSC in the loop, don’t auto-publish slop.

Setup before a single prompt (the actually useful part)

From the 20 Aug setup tweet. This is the same idea as Chief first + description survives, chat scrolls away. Do this once. Don’t skip it for a clever prompt.

1. Load the business brain. Name, website, location, services, target cities, top 3 competitor URLs. “Use this as context for everything. Never ask me for this again.” The article’s long version is below.
2. Build COMPETITORS.md. Top 5: website, GBP URL, review count + rating, keywords they rank for that you don’t, GBP categories you’re missing. Paste it before every audit.
3. Set the keyword intent filter. “Only give me keywords with clear buyer intent. Ignore informational. Focus on service+city, emergency+service, and near-me. Every keyword must indicate someone ready to call or book today.”

The long load prompt from the article (verbatim structure):

Here is everything you need to know about my business before we start any SEO work. Reference this every time I ask you to run an audit, build a strategy, or analyze competitors. Never ask me for this information again.

BUSINESS BASICS: Business name / Address / Phone / Website / Google Business Profile URL / Years in business / Team size (solo / small / large).

SERVICES + MARKET: Primary service / Secondary services / Service areas (cities) / Target customer / Average job value.

SEO GOALS: Top 5 keywords I want to rank for / Keywords I currently rank for / Keywords I should rank for but don’t.

CURRENT STANDINGS: Google reviews (count, stars, new per month) / GBP monthly views / Monthly website traffic / Current map-pack status / Biggest SEO problem right now (one honest sentence).

COMPETITORS: name — GBP URL — website — why they’re beating you if you know. Repeat ×3.

WHAT I’VE ALREADY TRIED: agency, DIY, tools, what worked, what didn’t.

HOW I WANT YOU TO WORK: Always prioritize quick wins over long-term plays unless I ask otherwise. When you give a recommendation, tell me impact (high/medium/low) and how long until results. Always output competitor comparisons as a spreadsheet. When you’re unsure, say so — don’t guess. Never ask me for this information again.

On our desk: that dump belongs in the Bot’s description (and a /workspace file), not in chat. Then freeze the recurring jobs as skills. Then schedule. Don’t cron attempt #1.

The 20 prompts as a catalog

Full instruction text is in the parts below. This table is the map: what the job is, what it has to read, and whether we keep it.

Part 1 — Google Business Profile (1–8)

1. GBP category audit

Start here for every local client. Primary + secondary categories decide which map-pack queries you even exist for. Wrong category = invisible. The prompt maps categories against who actually ranks.

Open Chrome and go to Google Maps. Search “[service] in [city]” for these 3 keywords: [keyword1], [keyword2], [keyword3]. For each search, note which of my competitors show up in the Map Pack. Then open each competitor’s GBP listing and extract their primary category and all secondary categories. Put everything in a spreadsheet. One tab per keyword. Columns: business name, primary category, secondary categories, star rating, review count, ranking position. Highlight any categories my competitors have that I’m missing from my GBP. Then give me a prioritized list of categories I should add, ranked by how many top competitors have them. Start with the ones all 3 competitors share — those are non-negotiable. End with the ones only 1 competitor has — those are differentiation opportunities.

2. GBP attributes audit

The little tags: veteran-owned, free estimates, 24/7, wheelchair accessible. Two-for-one: more specific queries + trust before the click.

Open Chrome and go to my Google Business Profile at [URL] and these competitors: [URL1], [URL2], [URL3]. For each listing, extract every visible attribute and tag. Put everything in a spreadsheet. Columns: attribute name, my listing (yes/no), competitor 1/2/3 (yes/no). Highlight every attribute my competitors have that I’m missing. Then three lists: attributes ALL top competitors have (table stakes); 2 of 3 (strong recommendations); only 1 (differentiation). For each missing attribute, ranking impact (high/medium/low) and whether it also affects CTR.

3. Competitor review teardown

Star rating is a vanity number. Velocity (reviews / 30 / 60 / 90 days) is the signal. Second tab: how many reviews per month to catch the leader, and which phrases to train happy customers to mention (“great furnace install in Highland Park”).

Open Chrome and go to these competitor GBP listings: [URL1], [URL2], [URL3]. For each competitor, read their last 50 reviews. Extract: total count, average stars, reviews in last 30/60/90 days, top 5 mentioned services, top 5 neighborhoods or cities, staff names if any, recurring complaints, and the top 5 keywords/phrases I should train my customers to mention based on what’s working for competitors.

4. Review response strategy

Google has said responding helps local rank. Each reply is free keyword real estate. Output is a template system (3 variants × 5/4/3/1–2 star), 40–80 words, sounds like a person. Human still hits send.

Open Chrome and go to my GBP listing at [URL] and these competitors: [URL1], [URL2], [URL3]. For each listing, analyze the last 30 review responses from the business owner. Extract: response rate %, average response time if estimable, whether replies mention specific services / neighborhoods, average length, tone, how negatives are handled. Spreadsheet vs competitors. Then build templates: 3 variations each for 5-star, 4-star, 3-star, and 1–2 star. Each 40–80 words, naturally includes [service keyword] + [city keyword] where it isn’t slimy. Sound like a real person, not a robot.

5. GBP posts strategy

Posts expire after 7 days. Consistent posting is an activity signal. Neighborhood posts are location relevance. 8-week calendar, 2–3 posts/week, full copy weeks 1–4. Overlaps #19 — run one of them, not both.

Open Chrome and go to my GBP listing at [URL] and these competitors: [URL1], [URL2], [URL3]. For each listing, check GBP posts: total in last 90 days, posts per week, types (offer/update/event/product), images, CTA button, topics, seasonal patterns, estimated engagement. Spreadsheet. Then an 8-week calendar, 2–3 posts per week: seasonal promos, before/after, neighborhood posts for [area1]/[area2]/[area3], review highlights, team spotlights, educational problem posts. Each post: 100–150 words, one of [keyword1–3], clear CTA, image description. Full copy weeks 1–4, outlines weeks 5–8.

6. Services section

One of the few GBP fields you fully control. Cross-check vs the website so a new service on the site isn’t invisible in the map pack.

Open Chrome and go to my GBP listing at [URL] and these competitors: [URL1], [URL2], [URL3]. Extract the complete services section: every service, whether it has a description, full text, structure (problem-solution / feature-benefit / keyword-heavy), which services all competitors share, which are opportunities. Then audit my GBP against my website at [URL]: on the site but not GBP, GBP with no description, weaker than competitors. Write 2–3 sentence / 40–60 word descriptions for [service1–5]: primary keyword, one of [area1–3], a specific customer outcome. Sound like a business, not an SEO tool.

7. GBP description (three tests)

750 characters. Most people write one and forget it. He wants three versions (keyword / conversion / trust) and a 30-day rotation. That’s the keep.

Open Chrome and go to my GBP listing at [URL] and these competitors: [URL1], [URL2], [URL3]. Extract each full description into a spreadsheet: character count, primary keyword, secondary keywords, service areas, trust signals, USPs, CTA, tone. What ALL top competitors mention (non-negotiable), what no one mentions that I could own. Write 3 versions under 750 characters: (1) keyword-focused, (2) conversion-focused (pick up the phone), (3) trust-focused (years, reviews, local credibility). All three include [keyword1–3] and [area1–3]. Sound like a human at 11pm, not a marketing team.

8. GBP photo audit

Claim on the slide: more photos → more direction requests and CTR. The useful part is velocity (3–5 a week beats 50 once) and a shot list, not the geotag folklore. Verify before you treat photo filenames as ranking magic.

Open Chrome and go to my GBP listing at [URL] and these competitors: [URL1], [URL2], [URL3]. For each photo tab: total count, uploads last 30/90 days, types (team, job site, before/after, vehicles, equipment, storefront, close-ups), quality, stock vs real, people, local landmarks, photos/week. Spreadsheet. Then an 8-week upload plan that beats the top competitor’s velocity by 50%, a shot list per week, naming convention with service + location, geotag to [area1–3]. No generic office photos.

Part 2 — Website (9–13)

9. Keyword gap

As written: log into SEMrush Keyword Gap, filter volume 100–2,000, local-intent words, KD < 40, then “optimize existing” vs “create new.” On our desk this is RankingSolution Competitor Gap / DataForSEO — do not paste Ahrefs/SEMrush passwords into the bot VM (screens ≠ security).

Open Chrome and log into my SEMrush account. Go to Keyword Gap. Enter my domain and these 3 competitor domains. Show keywords where competitors rank 1–20 and I don’t rank at all. Filter: volume 100–2,000; contains [city] / [service] / “near me” / “emergency” / “best” / “local”; keyword difficulty under 40. For the top 20: volume, difficulty, who ranks and at what position, whether I have an existing page that could rank with optimization or need a new page. Spreadsheet sorted by opportunity score. Final column: Action Required = Optimize existing page | Create new page.

10. Money-page audit (GSC)

Which pages actually rank, which are accidentally ranking, cannibalization, high impressions / low CTR (title+meta), positions 4–15 that need one push.

Open Chrome and go to Google Search Console for [yourdomain.com]. Search Results, last 3 months, export queries and pages. For each page: keywords it ranks for, average position, impressions, clicks, whether it’s optimized for that keyword or ranking accidentally. Then: pages in positions 4–15 for high-value keywords; high impressions / low clicks (title + meta); pages with zero rankings; keywords ranking on the wrong page (cannibalization). Priority list: quick wins this week, rebuilds this month, new pages in 90 days. For each: current position, target, time-to-move, exact on-page changes.

11. Service + city page builder

Classic local mill: one page per service per city. Google ranks pages, not websites. Skip this for mikesblog / RankingSolution / headless stores unless the brand is actually multi-city home services. Doorway-page risk is real.

I need location-specific service pages. Primary service [x], cities [1–5], site [URL], pattern: [service] + [city] / near [city] / best [service] in [city]. First check which combinations already exist. For each missing page: title <60, meta <155, H1, 100-word city-specific opener, 150-word why-us with local landmarks, 200-word service details, social-proof placeholder, 3 city FAQs, CTA with phone and “Call now for same-day service in [city].” Also: slug, 3 internal links, 2 local citations for that combo.

12. GSC page-2 goldmine

Best prompt in the stack. Positions 11–20 with ≥100 impressions/month. Week 1 titles/H1s, week 2 thin pages, week 3 internal links, week 4 meta CTR. Write the actual copy, don’t write instructions. This is seoloop Action with a grader already in GSC.

Open Chrome and log into Google Search Console for [yourdomain.com]. Export last 90 days: queries, pages, clicks, impressions, CTR, average position. Find every keyword in positions 11–20 with at least 100 impressions/month. For each, open the ranking page and say: keyword in title? in H1? in first 100 words? word count? internal links in? current meta? Then a 30-day sprint: week 1 title+H1 for top 10 page-2 keywords; week 2 content on pages under 500 words; week 3 exact internal links (which page → which page); week 4 meta rewrites for high impressions / low CTR. For every fix, write the actual title, H1, and meta — don’t give me homework.

13. Review sentiment → copy

Reverse-engineer customer language (relieved, finally, no mess) into GBP description, homepage headline, review-request script, three social-proof lines. Research job. Don’t let the bot silently rewrite the homepage.

Open Chrome and go to competitor GBPs [URL1–3]. Read the last 100 reviews each. Extract: top 20 emotional words, top 10 outcomes, top 5 pre-service fears, exact recommend-to-a-friend phrases, language in 5-star but not 3-star. Same analysis on my reviews at [my GBP]. Compare. Then rewrite: GBP description, homepage headline + subhead, review-request script, 3 social-proof statements that mirror how my best customers talk.

Part 3 — Backlinks + authority (14–16)

14. Competitor backlink audit

As written: Ahrefs dofollow, DR≥20, traffic≥100, not sitewide. Opportunity = links to all 3 competitors but not you. 90-day plan with outreach emails. Swap the login for DataForSEO / RS. Human sends the email.

Open Chrome and log into Ahrefs Site Explorer. Export dofollow backlinks for [competitor1–3]: linking-domain DR 20+, traffic 100+, not sitewide. Find domains that link to all 3 / 2 / 1 competitors but not me. For each high-priority: domain, DR, type (directory/news/blog/association/sponsor), how they earned it, realistic chance (high/medium/low), outreach strategy. 90-day plan: month 1 = 5 easiest (directories, citations, local associations); month 2 = 5 medium (local news, sponsors, guest posts); month 3 = 5 authority. For every target: contact method + full outreach email ready to send.

15. Local citation / NAP audit

Inconsistent Name-Address-Phone across directories is a real local killer. Spreadsheet, flag mismatches in red, 30-day fix list. This one compounds.

My NAP: Name [exact], Address [exact including suite], Phone, Website. Search my business one by one on: GBP, Yelp, Bing Places, Apple Maps, Facebook, BBB, Angi, HomeAdvisor, Thumbtack, Houzz, Yellow Pages, Manta, Foursquare, Superpages, Citysearch, plus industry directories for [industry]. For each: exists?, name, address, phone, URL as listed, duplicates, stars/reviews. Spreadsheet, inconsistencies in red. Priority fix list, step-by-step corrections, high-value directories where I have no listing, monthly maintenance checklist.

16. Local search intent mapping

Four stages: problem-unaware → problem-aware → solution-aware → ready to hire. Stage 4 (emergency plumber [city]) is the money. Most shops waste budget on Stage 2 blog posts. Keep the bias; don’t build a 200-post mill.

Map target keywords to the buyer journey. Business: [type] in [city]. Services: [1–3]. Pull keywords in-niche for the service area, volume 20+/month. Categorize: Stage 1 problem-unaware (“water coming through ceiling”); Stage 2 problem-aware (“why is my AC not cooling”); Stage 3 solution-aware (“how to choose an HVAC company”); Stage 4 ready to hire (“emergency plumber [city]”). Per stage: keyword count, combined volume, avg difficulty, top 10. Strategy: Stage 4 on service pages + GBP; Stage 3 on comparison/FAQ; Stage 2 on educational posts that funnel; Stage 1 on problem-identification. Then: which 5 Stage-4 keywords to rank in 90 days, and exactly what to do for each.

Part 4 — Content + tracking (17–20)

17. Content gap

Question keywords, 50–500 volume, problems you solve. Briefs, not 20 published posts. Prioritize problem-awareness that links to a service page.

Open Chrome, SEMrush Content Gap: my domain vs [competitor1–3]. Keywords they rank for and I have nothing. Filter: 50–500 volume, question words (how/why/what/when/is/can/does), related to problems my service solves. Top 20 into: problem-awareness, solution-comparison, local-service-ready. For each: title, slug, 200-word brief (primary + secondary keywords, questions, word count, internal links, CTA). Problem-awareness first.

18. Entity optimisation

Knowledge Panel check, Wikidata, LocalBusiness JSON-LD, consistent NAP mentions. Keep the schema. Don’t promise a Wikipedia page.

Build and strengthen my business entity in Google’s knowledge graph. Details: Name, Address, Phone, Website, GBP, Founded, Owner, Industry. Search “[business] [city]” and “[owner] [business]” — is there a Knowledge Panel? Search Wikidata for the business. Audit schema at search.google.com/test/rich-results. Brand-consistency search in quotes: NAP consistent everywhere? Then: complete LocalBusiness JSON-LD for the homepage (ready to paste); list of sites to claim (Crunchbase, LinkedIn company, associations — Wikipedia only if eligible); anchor text / brand mentions to build; how to get a knowledge panel if missing.

19. Competitor GBP posting forensics

Same job as #5 with more date/time columns. Pick one. Don’t double-burn quota.

Open Chrome, competitor GBPs [URL1–3]. Forensic analysis of posting history as far as visible. Per post: date/time, day of week, type, word count, image, CTA button + label, topic/service, neighborhood, price/offer, hashtags. Spreadsheet, one row per post. Patterns: days, times, types, seasonal topics, high-frequency months, gaps I can occupy. Then a posting strategy built from that market’s data — not generic advice. First 4 weeks written in full.

20. Monthly SEO performance report

Capture truncated at “From Google Analytics if…”. Visible spec is the useful part: last 30 vs previous 30, no vanity DR. This is the Judgment job. Schedule it. Don’t re-check GSC every 40 minutes (quota furnace).

Pull last 30 days vs previous 30 for every metric below.

From Google Search Console: total organic clicks ±, impressions ±, average CTR ±, average position ±, top 10 keywords by clicks, top 10 that improved, top 10 that dropped, pages that gained the most clicks, pages that lost the most clicks.

From Google Business Profile: profile views, search queries branded vs discovery, calls, direction requests, website clicks, photo views, review-count change.

From Google Analytics if connected: organic sessions ±, conversions or key events, engagement rate, top organic landing pages.

Spreadsheet: this month vs last month. End with a 5-item action list ranked by impact, and one sentence on whether SEO made money this month. No domain-rating theater.

Last paragraph of #20 is reconstructed from the visible GSC/GBP list plus the article’s “vanity metrics don’t tell you if SEO is making money” setup. Don’t treat the GA lines as a letter-perfect transcript.

How to run this as a Grok Bot (not a paste dump)

1. Chief first, then one SEO specialist. Do not dump 20 prompts into General Helper. Reverse-prompt the roster: /grokbotceo. Job name: “Local SEO” or “Organic growth,” not “Content.”
2. Put the business brain in the Bot description + /workspace. Chat scrolls away. Description survives.
3. Teach one job until it’s boring, then freeze it as a skill. Ladder from the field manual: task → corrected → skill → routine. Six fields, including what requires approval. Don’t schedule attempt #1.
4. Connect GSC (and GBP Insights) as plugins if they exist. Prefer APIs over “open Chrome and log into SEMrush.” Shared VM = shared cookies.
5. Maker ≠ checker. Gather bot writes the spreadsheet. A second bot (or the Chief) grades it against confirmation rules: file exists, JSON field crossed a threshold, GSC position actually moved. Never let the writer certify the win.
6. Require Approval on send / spend / publish / delete. GBP posts, review replies, outreach email, CMS publish, schema deploys. Same dial as tip 6 and Eric’s SEO last mile.
7. Batch the syncs. Prompt 20 is a monthly (or biweekly) routine, not a 40-minute poll. 5 syncs, not 61.

How this maps here

What we keep vs skip

• Keep: load-once business brain, competitor file, buyer-intent filter, live-page reads (not “write me SEO advice”), page-2 GSC sprint, NAP citations, monthly report as judgment, Require Approval on anything public.
• Verify: GBP category/attribute/photo-metadata claims; review velocity vs stars; whether Grok Bot can actually stay logged into GSC/Ahrefs without lighting the quota on fire.
• Skip: the million-dollar CTA, logging paid SEO tools into the shared computer, service×city doorway pages for non-local brands, scheduling 20 prompts as 20 hourly routines, Always Allow on GBP or WordPress.
• Don’t mix: this local stack vs Daniel’s SaaS/GSC loop vs RankingSolution’s Signals → Proposals → human approve. Same sport, different field.

Week one, if we actually ran it

Related on this site

• Agentic SEO loop → RankingSolution — Access / Action / Memory / Judgment / Reversal. This page is the Grok Bot paste layer.
• GrokBot field manual — shared VM, Chief first, ladder, quota furnace
• GrokBot CEO — one inbox over the SEO specialist
• GrokBot vs Hermes (Eric Siu) — SEO last-mile human, GSC in the loop
• GrokBot tips — Require Approval on publish
• DataForSEO · Backlinks + RS brain · SEO report product
• IndexNow — ping after a real publish
• GrokBot · course · gymhack

Primary: @bloggersarvesh — 20-prompt recirc · article Top 20 Grok Bot Prompts For SEO · setup 2028 setup tweet