Run all of this inside the full Agent OS — join the AI Profit Boardroom →

agentos.guide › blog

AI SEO Digital Agency Groundwork: The Five Agent OS Prompts To Run Before You Brief One

By Julian Goldie · 11 November 2026 · agentos.guide

Illustrated agent console checking a robots file, a web page wireframe, schema brackets and a brief before passing a folder to a team, on a dark aubergine background with gold accents

Before you hire an AI SEO digital agency, run the groundwork through your Agent OS so you know exactly what to ask for.

Five copy-paste prompts check crawler access, audit your new build, draft schema, brief a service page and research link targets, all read-only or draft-only.

Then you hand the agency a clear list rather than a vague wish. When you want a human on it, book a free SEO strategy session.

Quick start: set up the folder in five minutes

  1. Create a folder called ai-seo-groundwork in your Agent OS workspace.
  2. Add a rules file that says: read-only on the website, draft-only for anything outbound, never log in to the CMS.
  3. Paste in your sitemap URL, your Google Business Profile details and a list of your top ten pages.
  4. Export the last 90 days from Search Console, including the generative AI performance report, and drop the CSV in.
  5. Run the prompts below in order and save each output as its own file.

If you have not installed an Agent OS yet, the AI Profit Boardroom includes it with setup trainings.

Google says AI Overviews and AI Mode need no special optimisation beyond normal SEO, so this groundwork is ordinary SEO hygiene aimed at the places AI tools read from.

Why the groundwork splits into web, content and local

A digital agency usually splits its people into three teams, and AI search touches each of them.

The web team controls whether crawlers can reach and render your pages.

The content team controls whether those pages say anything worth quoting.

The local team controls whether your Business Profile and listings tell the same story as your site.

The prompts below map to those three teams, so each output file lands on the right desk.

Bing makes the same point from its side: Copilot runs on the same crawling and ranking foundation as classic search, so fixing the basics helps everywhere at once.

Prompt 1: check AI crawler access, including the CDN

OpenAI only shows sites in ChatGPT search if OAI-SearchBot is allowed and the host or CDN lets its IP ranges through.

Perplexity and Anthropic publish similar crawler lists in their Perplexity crawler docs and Claude crawler help page.

You are my AI search access checker. Read only. Do not change any files.
Site: https://www.example.co.uk

1. Fetch /robots.txt and list the rules that apply to:
   Googlebot, Bingbot, OAI-SearchBot, GPTBot, PerplexityBot,
   Claude-SearchBot, ClaudeBot, Google-Extended.
2. For each bot, say ALLOWED or BLOCKED for the home page,
   one service page and one location page.
3. Request each of those three URLs and report the HTTP status.
   Flag any 403, 429 or challenge page, as that usually means a
   CDN or firewall rule.
4. Mark which bots are search crawlers and which are training crawlers.
5. Output a table plus a short list of fixes for the agency's developer.

Search crawlers and training crawlers are separate switches. Blocking GPTBot or ClaudeBot is a business choice that does not remove you from ChatGPT or Claude search.

Prompt 2: audit the new build before launch day

Digital agencies often build on JavaScript frameworks. Google renders JavaScript but still calls server-side or pre-rendering a great idea, because not every bot can run it.

The same audit catches stray nosnippet and max-snippet rules, which also limit how your content is used in AI Overviews.

Audit these 10 URLs from my new website. Read only.
[paste URLs]

For each URL report:
- HTTP status and canonical URL
- any meta robots or X-Robots-Tag values (noindex, nosnippet, max-snippet)
- any data-nosnippet sections and what they wrap
- whether the main heading and first 200 words appear in the raw HTML
  before JavaScript runs
- internal links that are not plain <a href> links
Finish with the three highest-risk problems, in plain English,
for a non-technical owner to forward to the web team.

🔥 Prompts flagging more than you can fix?

Book a free SEO strategy session and bring the crawler report and render check your agent produced.

→ Reserve your strategy session

Prompt 3: draft LocalBusiness schema that matches the page

Google says there is no special AI schema, but LocalBusiness markup still helps it understand who you are and where you trade.

The key rule is consistency: the name, address and phone must match your Business Profile and your visible page.

Draft LocalBusiness JSON-LD for my business. Draft only, I will check it.
Use ONLY facts from these pages: [contact page URL], [about page URL].
Include: @type (pick the most specific LocalBusiness subtype),
name exactly as on my Google Business Profile, url, telephone,
address, openingHoursSpecification, and sameAs links to my real
profiles. Leave out anything you cannot find on the pages and list
what is missing. Do not add reviews or ratings.

Paste the output into Google's Rich Results Test, then give the cleaned version to the agency's developer.

Prompt 4: brief a service page from your own customers

Google's optimisation guide asks for non-commodity content: first-hand detail that a summary of other sites cannot match.

Your call notes and customer emails are the best raw material you own.

Build a content brief for a service page on [service] in [town].
Inputs: these call notes and customer emails [paste], my current
page [URL] and the top 5 results for "[service] [town]".
Give me:
- the 8 questions customers actually ask, in their words
- what my page answers and what it misses
- first-hand details I should add (photos, steps, timings, examples)
- a direct two-sentence answer for the top of the page
Write in UK English. No keyword stuffing. Do not invent facts.

Hand the brief to the agency's writer. You stay in charge of the facts; they handle the craft.

Prompt 5: research link targets without sending a thing

Links still matter for discovery and authority, and they must stay inside Google's link spam policy.

Paid placements need a sponsored label. Earned mentions from real local and industry sites are the goal.

Research link and mention opportunities for my business. Draft only.
Do NOT send any email or fill in any form.
Business: [one-line description], area: [region].
Find 20 real local or industry sites (associations, local news,
suppliers, event pages) that already link to similar businesses.
For each: URL, why it fits, the contact page, and a 3-line pitch draft.
Skip anything that sells links without a sponsored label, link farms
and private networks. Save as a table for me to review.

🧠 Need the Agent OS these prompts run in?

Inside the AI Profit Boardroom you get the installable Agent OS, step-by-step trainings and four coaching calls a week, so you can learn AI SEO yourself.

→ Join the AI Profit Boardroom

A short workflow for the first month

WeekWhat your agent doesWhat you do
1Runs Prompts 1 and 2Sends the fix list to the web team
2Drafts schema with Prompt 3Checks every fact, then approves
3Builds two briefs with Prompt 4Adds photos and first-hand detail
4Researches targets with Prompt 5Reviews the list and decides who to contact

Re-run Prompt 1 every week. A single CDN update can quietly block a crawler again. Track Copilot citations in Bing Webmaster Tools AI Performance alongside Search Console.

Worked example: a family-run garden centre

  1. The owner runs Prompt 1 and finds the CDN returns a challenge page to PerplexityBot and OAI-SearchBot.
  2. Prompt 2 shows the new plant-care pages only load their text after JavaScript runs.
  3. Prompt 3 drafts GardenStore schema from the contact page, and she corrects the Sunday hours.
  4. Prompt 4 turns staff notes on winter pruning into a brief with eight real customer questions.
  5. She sends the agency all four files with one line: fix access and rendering first, then write from the brief.

The whole exercise took her an afternoon spread over two weeks.

The agency started on the right problems on day one instead of spending its first fortnight finding them.

Six weeks later she compared AI feature impressions in Search Console and chatgpt.com referrals in analytics with the export she saved in step four of the quick start, and logged the phone enquiries alongside them.

What to ask the agency once your files are ready

A good agency will welcome the files. A defensive reaction to a clear brief tells you something useful too.

When to hand it to a pro

Your Agent OS does the checking and the first drafts. It cannot negotiate editorial links, decide strategy for a competitive market or own a migration.

A digital agency is the right partner for the build, content and local work in your files.

A specialist SEO agency fits better when the gap is authority, technical recovery or a crowded niche.

Goldie Agency works on that specialist side: Julian Goldie's own published numbers list 1,000+ SEO and link building campaigns and 240+ clients, with link building done by manual outreach, no PBNs.

Nobody, specialist or not, can guarantee a #1 ranking or a ChatGPT mention, and Google says so plainly.

For the specialist view of this work, read the AI SEO specialist guide.

🔥 Groundwork finished? Hand it over properly.

Book a free SEO strategy session and leave with a split of what your digital agency should own and what needs a specialist.

→ Book your free session

🎁 Free extras for your prompt library

Load the 200+ AI SEO Prompts into your Agent OS and read the free link building book before you run Prompt 5.

→ Book a free SEO strategy session

Also On Our Network

FAQ

Can an Agent OS replace an AI SEO digital agency?

It can do the checks and first drafts. It cannot rebuild your site, earn editorial links or make strategy calls, which is where an agency earns its place.

Is it safe to point an agent at my website?

Yes, if the rules file keeps it read-only on the site and draft-only for anything outbound, with no CMS logins.

Which prompt should I run first?

Prompt 1. If AI search crawlers are blocked at robots.txt or the CDN, nothing else in the plan can get you into those answers.

Does my agency need to add llms.txt for Google?

No. Google says Search does not use llms.txt files, though other services may read them.

How often should I re-run the prompts?

Run Prompt 1 weekly and Prompt 2 after every site release. Briefs and link research can run monthly.

Final word on AI SEO Digital Agency Groundwork

Run the five prompts, keep the folder current, and book a free strategy session when you are ready to brief an AI SEO digital agency.

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.

→ Get my best AI training inside the AI Profit Boardroom

Related guides on agentos.guide

→ AI SEO Company Groundwork: 6 Prompts To Run In Your Agent OS

→ Hire SEO Agency Prep: Do The Groundwork In Your Agent OS

→ Outsource SEO Services: Do the Groundwork In Your Agent OS

→ AI SEO Firm Prep: 6 Prompts To Run In 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 👉

🆓 Get a FREE AI Course + Community + 1,000 AI Agents 👉

← all guides on agentos.guide