agentos.guide › blog
LLM SEO Services, Started In Your Agent OS: A Citation Baseline, Crawler Check and Four More Prompts
LLM SEO services start with groundwork your Agent OS can do this afternoon: a citation baseline, an AI crawler check, entity markup, one answer-page brief and a list of sites worth earning a mention on.
Do it yourself and you'll either fix the easy wins or arrive at any agency call with the evidence already gathered.
When the outputs are ready, book a free SEO strategy session and bring them along.
Quick start: set the agent up safely
Give your agent a working folder with three files before you paste a single prompt.
site.txt: every hostname you run, including blog, shop and help centre subdomains.business.txt: your exact business name, address, phone, services and the URLs of your official profiles.questions.txt: 20 questions a customer might type into ChatGPT or Perplexity before choosing a business like yours.
Set permissions to read and draft only.
The agent can fetch pages, read files and write drafts, but it shouldn't edit your live site, change DNS or send a single email.
Prompt 1: Build a citation baseline
Before changing anything, record who the AI tools cite today, so you can measure progress later.
Read questions.txt. For each question, run a web search the way an
AI answer engine would and list the top 8 sources you would cite.
Save a table to baseline.csv with columns:
question | cited domains | is our domain cited (yes/no) | competitor cited first
Then summarise: how many questions cite us, and which 5 domains
appear most often. Do not contact any site.This is a proxy, not the real engines, so also spot-check five questions by hand in ChatGPT, Perplexity and Google's AI Mode.
Note the date on every file; you'll repeat this monthly.
Prompt 2: Check AI crawler access and preview controls
Search crawlers and training crawlers are separate, and each company documents its own: OpenAI, Perplexity and Anthropic.
For every host in site.txt:
1. Fetch /robots.txt and report, for each user agent, allowed or blocked on / :
OAI-SearchBot, PerplexityBot, Claude-SearchBot, Googlebot, bingbot (search)
GPTBot, ClaudeBot, Google-Extended (training)
2. Fetch the homepage and 5 service pages. Report any meta robots or
X-Robots-Tag values containing noindex, nosnippet, max-snippet,
noarchive or nocache.
3. Flag in red any SEARCH crawler that is blocked, and any money page
with nosnippet or noarchive.
Read only. Change nothing. Output one table per host.Why the snippet tags matter: Google says nosnippet stops content being a direct input for AI Overviews and AI Mode, and Bing says NOARCHIVE keeps content out of Copilot responses.
And don't panic about Google-Extended: Google's common crawlers page says it has no effect on inclusion in Google Search and isn't a ranking signal.
If robots.txt looks fine but you suspect a block, ask your host whether a firewall or CDN bot setting challenges these bots, as OpenAI and Perplexity both ask hosts to allow their published IP ranges.
📞 Found blocked bots or odd tags?
Send the tables to Julian's team on a free call. Book a free SEO strategy session and get the fixes ranked in order.
Prompt 3: Draft Organization schema with sameAs
Clear entity data helps engines work out which business you are; Google's Organization structured data guide says this markup helps it disambiguate your organisation.
Using business.txt only, draft JSON-LD for schema.org Organization
(or LocalBusiness if we have a public address). Include name, url,
logo, telephone, address, and sameAs with each official profile URL.
Use only facts in business.txt; mark anything missing as TODO.
Then compare the name, address and phone against our homepage footer
and list every mismatch. Output the JSON-LD and the mismatch list.
Draft only; do not publish.Keep it accurate: Google says there's no special schema needed for its AI features, and markup must match what's visible on the page.
If your agent suggests an llms.txt file, that's fine for other systems, but Google's guide says Search doesn't use it.
Prompt 4: Brief one answer page
Pick the question from your baseline where competitors are cited and you aren't, then brief a page that answers it better. Google's generative AI guide favours non-commodity content: first-hand detail others can't copy.
Take question #[N] from baseline.csv.
Read the 3 pages cited for it and list what each covers and misses.
Write a brief for one page on our site that answers the question
directly in the first two sentences, then adds what only we can say:
our process, real examples, photos to take, numbers we can verify.
Include: suggested title, H2 outline, facts we must confirm, and the
internal pages that should link to it. Do not write the page itself.Write the page yourself from the brief, or hand the brief to whoever writes for you.
One strong page per question beats ten thin ones; Google calls mass-produced variations scaled content abuse.
Prompt 5: Research a mention list, draft only
AI answers draw on what the wider web says about you, so a short list of the right sites is worth more than a long list of any sites.
From baseline.csv, take the 15 domains cited most often.
For each, note: site type (directory, publication, review site,
forum, association), whether businesses like ours appear there,
and the legitimate route in (listing form, editorial pitch, review
request, membership). Skip any site that sells links that pass
ranking credit. Draft a one-paragraph pitch for the top 5.
Draft only. Do not send anything or fill in any form.Paid placements are allowed only when marked rel="sponsored" or nofollow, per Google's spam policies; editorial mentions earned by outreach are the goal.
🔥 Learn AI SEO yourself, inside the AI Profit Boardroom
Get the installable Agent OS, step-by-step AI SEO trainings and four live coaching calls a week where you can check these prompts on your own site.
A worked example: a local physio clinic
Step 1: the clinic's agent runs Prompt 1 and finds it's cited for 2 of 20 questions.
Step 2: Prompt 2 shows the booking subdomain blocks every bot with one wildcard rule.
Step 3: Prompt 3 flags three different phone numbers across the site and profiles.
Step 4: Prompt 4 briefs a page on recovery timelines for a common injury, written by the lead physio with real case notes, kept accurate to clinical guidance.
Step 5: Prompt 5 lists the local health directories and the sports club that keep getting cited.
None of that needed an agency, and all of it makes an agency's job faster if you hire one.
Mistakes to avoid when you run these prompts
The most common slip is trusting the agent's proxy search as if it were ChatGPT itself; it's a sample, so treat movement over months as the signal.
The second is letting the agent tidy robots.txt on its own, which can lock out a search bot in one keystroke.
The third is publishing the schema draft with TODO fields still in it, or with an address that doesn't match the footer.
The fourth is turning Prompt 4 into ten near-identical pages, which is exactly the pattern Google warns against.
And the last is pitching every domain on the list, including ones that only sell links; skip those and keep the pitch list short and relevant.
If you want a second pair of eyes on your outputs, members of the AI Profit Boardroom bring them to the weekly coaching calls.
The monthly loop
- Re-run Prompt 1 and compare with last month's baseline.
- Re-run Prompt 2 after any site, host or CDN change.
- Check Search Console's generative AI report and Bing's AI Performance report.
- Brief one new answer page and pitch five mentions.
Bing's AI Performance report shows citations and the grounding queries behind them across Copilot, which is the closest first-party view of what the agent's proxy is trying to estimate.
In our own analytics, ChatGPT visits usually appear under chatgpt.com as the source, which makes a handy filter alongside the reports.
When to hand it to a pro
Hand it over when the groundwork shows problems your agent can only describe, not fix.
- Bots blocked at the firewall and nobody knows who manages the CDN.
- Pages that come back empty without JavaScript.
- Competitors cited everywhere and you have no mentions to show.
- No time to write the answer pages each month.
No one can promise a ChatGPT citation; OpenAI says placement isn't guaranteed, and Google says no one can guarantee a #1 ranking.
What a good team can do is the work behind it, and Goldie Agency's own published numbers point to 1,000+ SEO and link building campaigns and 240+ clients.
📞 Hand over the groundwork
Bring your baseline, crawler tables and briefs to a free call. Book a free SEO strategy session and leave with a plan.
🎁 Two free resources
Grab the 200+ AI SEO Prompts to extend this list and the free link building book for Prompt 5. Or learn it all in the AI Profit Boardroom.
Also On Our Network
FAQ
Can my agent check what ChatGPT really cites?
Only roughly. Its web search is a proxy, so spot-check a handful of questions by hand in ChatGPT, Perplexity and Google's AI Mode.
Is it safe to let the agent edit robots.txt?
Keep it on read and draft. Let it propose changes, then have a person review and deploy them.
Which bots should stay allowed?
The search ones: OAI-SearchBot, PerplexityBot, Claude-SearchBot, Googlebot and bingbot. Training bots are a separate decision for you.
How often should I re-run the prompts?
Monthly for the baseline and briefs, and straight after any site, host or CDN change for the crawler check.
Final word on LLM SEO Services, Started In Your Agent OS
Run the five prompts this week, keep every output as a draft, and book a free strategy session when you're ready to hand the work to LLM SEO services.
About Julian
I'm Julian Goldie, an AI entrepreneur, SEO expert and the founder of the AI Profit Boardroom, which has 3,400+ members.
I help business owners scale with AI agents, automation and SEO.
My YouTube channel has 400,000+ subscribers, and I run Goldie Agency, a seven-figure SEO agency.
Related guides on agentos.guide
→ Generative AI SEO Services: Prep 5 Of The 7 In Your Agent OS
→ LLM SEO Agency for SaaS Groundwork: 5 Agent OS Prompts
→ How To Rank In ChatGPT: 5 Prompts For Your Agent OS
→ Best ChatGPT SEO Agency: Vet Them With Your Agent OS
Get the whole Agent OS
The installable system behind every guide on this site — dashboard, agents, memory and pipelines — updated as new tools land, with 3,400+ members and four coaching calls a week.
Join the AI Profit Boardroom →📺 Video notes + links to the tools 👉
🎥 Learn how I make these videos 👉