agentos.guide › blog

SEO Agency For SaaS Groundwork In Your Agent OS: 5 Prompts To Run Before You Hire

By Julian Goldie · 11 November 2026 · agentos.guide

Before you pay an SEO agency for SaaS to audit your software site, your Agent OS can do the groundwork in an afternoon.

It can compare raw and rendered pages, check robots.txt for AI search crawlers, draft schema, and research outreach targets.

Below is a quick start, five copy-paste prompts and a clear line for when to bring in a team.

If you would rather skip straight to a plan, book a free strategy session with Julian's team.

Quick Start: Set Up The SaaS SEO Folder

1. Create a folder called saas-seo in your Agent OS workspace.
2. Add urls.txt with one URL per template: home, a feature page, an integration page, pricing, a docs page.
3. Add brand-facts.md with your exact company name, product name, logo URL and official profile links.
4. Export recent support tickets to support-tickets.csv and paste sales notes into sales-notes.md.
5. Turn on the setting that makes the agent ask before any write, send or publish step.

Every prompt below is draft-only, so that last step is your safety net.

Date-stamp each output file so you can compare runs month to month.

Give The Agent The Rules First

Save Google's JavaScript SEO basics page into the folder.

It explains that Google crawls, renders, then indexes, and only sees what is in the rendered HTML.

Add OpenAI's crawler overview, which says OAI-SearchBot and GPTBot are independent settings.

Add Google's AI optimisation guide as well; it tells site owners that llms.txt and similar AI text files are not needed for Google Search.

Tell the agent to cite these files instead of guessing.

The Five Copy-Paste Prompts

Prompt 1: Raw Versus Rendered Audit

You are my SaaS rendering auditor. Draft only, change nothing.
For each URL in urls.txt:
1. Fetch the raw HTML (no JavaScript) and save the visible text.
2. Load the page in a headless browser and save the rendered text.
3. Report: H1 present in raw? H1 present in rendered? word count raw vs rendered.
4. Flag any <meta name="robots" content="noindex"> in the raw HTML.
5. Flag links that use "#/" fragments or have no href.
6. Note the canonical in raw HTML and whether JavaScript changes it.
Write rendering-audit.csv, then list the five templates with the biggest gap.

This catches the classic SaaS problem: a marketing page that is an empty app shell until scripts run.

Google warns that a noindex in the original code may stop rendering entirely, so removing it later with JavaScript may not work.

Prompt 2: Robots.txt Check For AI Search Bots

Act as my AI crawler checker. Draft only.
Fetch https://MYDOMAIN/robots.txt.
For Googlebot, Bingbot, OAI-SearchBot, GPTBot, PerplexityBot,
Claude-SearchBot and ClaudeBot, report which group applies and whether
/, /pricing, /docs/ and /integrations/ are allowed.
Flag any "User-agent: *" rule that blocks a search bot by accident.
Then request the homepage with each search bot's user-agent string and
record the HTTP status, so I can spot firewall or CDN blocks.
Propose a corrected robots.txt as a diff. Do not apply it.

PerplexityBot comes from Perplexity’s crawler docs, and Claude-SearchBot from Anthropic’s crawler help page.

Both say blocking their search crawler reduces your visibility in their results.

OpenAI's ChatGPT search help article adds that your CDN must allow its published searchbot IP addresses.

🔥 Audit done? Get a second opinion on the fixes

Bring your rendering-audit.csv to a call and Julian's team will help you rank the fixes for your developer. Book a free SEO strategy session.

→ Get the free SaaS SEO plan

Prompts For Entities And Content

Prompt 3: Organization Schema And Brand Consistency

Draft Organization JSON-LD for my homepage. Draft only.
Use ONLY the facts in brand-facts.md (name, URL, logo, profile links).
Include a sameAs array with every official profile listed there.
Then compare the brand name, product name and tagline across the homepage,
docs home, app store listing and LinkedIn page, and list every mismatch.
Mark any field you cannot source as [NEEDS SOURCE].

Google's Organization markup docs say it helps Google tell your organisation apart from others with similar names.

That matters for software, where product names often overlap.

Don't expect it to unlock AI Overviews, though: Google says no special schema is needed for its AI features.

Prompt 4: Brief A Comparison Page

Write a content brief for a comparison page: OUR TOOL vs COMPETITOR.
Inputs: sales-notes.md, feature-list.md, support-tickets.csv.
1. List the five questions prospects ask most about this comparison.
2. Build a feature table using only facts from feature-list.md.
3. Flag any claim about the competitor that needs a public source link.
4. Suggest one first-hand detail (screenshot, workflow, customer quote
   we have permission to use) per section.
Output a brief, not the article. No invented numbers.

Comparison and alternatives pages catch buyers who are already switching.

Keep them fair and factual, and let your own product knowledge do the work.

Google's guide asks for non-commodity content, and first-hand screenshots are exactly that.

Prompt 5: Draft An Outreach Target List

Research link and mention targets for my SaaS. Draft only, send nothing.
1. List integration partners from integrations.txt that run a public
   partner directory or app marketplace.
2. Find industry publications and podcasts in NICHE that covered
   comparable tools in the last year.
3. For each, note the page URL and why it is relevant.
4. Mark any target that only sells links as "skip".
Save outreach-targets.csv. A human writes and sends every pitch.

Google's spam policies allow paid placements only with rel="sponsored" or rel="nofollow".

Keep the agent on research, and keep a person on every pitch.

A Simple Weekly Workflow

1. Monday: run Prompt 1 on any template that changed last week.
2. Tuesday: run Prompt 2 after every deploy that touches robots.txt or the CDN.
3. Wednesday: run Prompt 3 whenever the brand name, logo or a profile link changes.
4. Thursday: pick one comparison or use-case page and run Prompt 4.
5. Friday: run Prompt 5 and pass the shortlist to whoever writes the pitches.

The whole loop takes about an hour of your time once the folder is set up.

The agent does the fetching and comparing; you make the calls.

Reading The Outputs Without Getting Lost

Start with rendering-audit.csv, because a page Google can't see makes every other fix pointless.

Sort it by the gap between raw and rendered word counts and fix the largest gaps first.

Next, read the robots.txt diff and check it against what each crawler's own documentation says.

Only then move to schema, briefs and outreach.

Track progress in Search Console's generative AI performance reports and in Bing's AI Performance report, which shows Copilot citations.

From what we see, ChatGPT referrals tend to show up in analytics with chatgpt.com as the source, so set up a segment for it.

Gotchas To Watch For

• Agents can misread a page behind a login or bot wall. Spot-check any odd result by hand.
• Some headless browsers render differently from Googlebot. Confirm big gaps with URL Inspection.
• Never let the agent publish a comparison page. Claims about competitors need a person and a source.
• Blocking Google-Extended does not change Google Search, according to Google's crawler docs, so don't expect a ranking effect from it.

A Small-Business Run-Through

A two-person team sells a stock-alert app for independent bike shops.

1. Prompt 1 shows the integrations page has 40 words before rendering and 600 after. The developer pre-renders it.
2. Prompt 2 finds Disallow: / under User-agent: * left over from staging, which also blocks OAI-SearchBot.
3. Prompt 3 spots that the app store listing uses an old product name. They fix it.
4. Prompt 4 produces a brief for a page comparing the app with spreadsheet stock tracking.
5. Prompt 5 lists two cycling trade magazines and a point-of-sale partner directory.

That is a week of agency discovery done in-house.

None of it needed a specialist tool, only a clear folder, the right source pages and an agent that asks before acting.

When To Hand It To A Pro

Hand it over when the fixes need sustained engineering time, when outreach needs relationships, or when nobody owns the plan.

Julian Goldie's agency, Goldie Agency, covers SEO, GEO, AEO and link building, and sums up its outreach in four words: manual outreach, no PBNs.

Julian's about page lists 1,000+ SEO and link building campaigns, and his channel has 425K+ YouTube subscribers.

Google's hiring advice says no one can guarantee a #1 ranking, so judge any team on its process.

Book a free SEO strategy session and bring your five output files.

For the AI-answer side, see the LLM SEO agency for SaaS guide.

🔥 Learn AI SEO yourself with the full Agent OS

The AI Profit Boardroom gives you the installable Agent OS, step-by-step trainings and four live coaching calls a week.

→ Join the AI Profit Boardroom

What To Bring To The Call

• rendering-audit.csv, sorted by the biggest raw-versus-rendered gap.
• The robots.txt diff from Prompt 2, plus the status codes each bot received.
• The list of brand-name mismatches from Prompt 3.
• One comparison brief from Prompt 4.
• The outreach shortlist from Prompt 5, with skipped targets marked.

With those five files, a strategy call skips discovery and goes straight to priorities.

You will know which fixes your developer can ship alone and which need specialist help.

You will also have a clear record of what the site looked like before any agency touched it, which makes later reporting accurate.

Free Extras For Your First Run

Grab the 200+ AI SEO Prompts pack and the free link building book, SEO Link Building Mastery.

To go further, learn AI SEO yourself inside the Boardroom, or Book a free SEO strategy session for a team plan.

Your Agent OS can do the groundwork, but a person should check the plan. Book a free SEO strategy session when you are ready.

FAQ

Can my Agent OS replace an SEO agency for SaaS?

It can handle discovery: rendering checks, robots.txt reviews, schema drafts and research. Engineering fixes, outreach and strategy still need people.

Why compare raw and rendered HTML?

Google indexes what appears after rendering, and some bots do not run JavaScript at all, so gaps between the two show what is at risk.

Should the agent change robots.txt itself?

No. Let it propose a diff and have a person apply it, then note the date so you can check crawl changes later.

Does llms.txt help my SaaS rank in Google?

Google says you can ignore AI text files like llms.txt for Google Search, so focus on rendering, access and content.

Final word on SEO Agency For SaaS Groundwork In Your Agent OS

Run the five prompts, fix what they find, and book a free strategy session when you are ready for an SEO agency for SaaS to take the rest.

Related guides on agentos.guide

→ LLM SEO Agency for SaaS Groundwork: 5 Agent OS Prompts

→ SEO Agency for Startups: Do the Groundwork in Your Agent OS

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

→ ChatGPT SEO Agency Groundwork: 5 Agent OS Prompts

📺 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