Guides published jul 23, 2026
In this guide, you will learn how to build a team of Gumloop agents that audits one page from your site and organizes the findings in one Google Sheets workbook. The agents will add backlink, content refresh, and technical SEO findings, then combine them into one Issues queue.
You will build a shared SEO audit workbook powered by a Backlink Agent, Content Refresh Agent, and Technical SEO Audit Agent. Each one reviews the same page and adds its findings to the workbook. A final Issues tab turns the recommendations into one prioritized queue for a teammate, contractor, or client.
Give your main SEO audit agent your website URL and tell it to coordinate the other agents, audit the first 25 pages it finds, and create the full workbook.
Run a full SEO audit for:
[WEBSITE URL]
Coordinate the Technical SEO Audit Agent, Backlink Agent, and Content Refresh Agent.
Audit the first 25 HTML pages you find on the same hostname. Use Firecrawl for public-web research and Google Sheets for the output.
Create one Google Sheets workbook with these tabs:
- Backlinks
- Content Refresh
- Technical SEO Audit
- Issues
Preserve source URLs and coverage limitations. Do not edit the website, publish changes, or send outreach.
Return the workbook URL, pages reviewed by each agent, failures, issue totals, and the biggest coverage limitation.Review the workbook and credit use before increasing the crawl limit. If you target older blogs for content refreshes, make sure each page exposes a published date that Firecrawl can read. Keep a human review step before anyone edits a page, publishes content, or sends outreach.
Open Gumloop's Backlink Specialist template (https://www.gumloop.com/templates/backlink-specialist) and select Use this Agent. The template gives you a fast starting point for the internal-link job: inspect one page, map the same site, and suggest relevant pages to connect.
Rename the duplicate Backlink Agent. In Connectors, keep Firecrawl and add Google Sheets. You do not need the template's Google Docs connector for this workbook workflow.
Use Firecrawl Scrape for the target page and every recommended destination, Map to discover pages on the same site, and Google Sheets to create the workbook and write the findings. Leave email sending, CMS publishing, page editing, and delete tools off.
Choose one page from your site that you want to audit and copy the URL.
Pro tip: A blog that needs a content refresh or a product page that needs better keywords is a great place to start. Testing one page also lets you review the output and credit use before scaling.
Give the Backlink Agent the URL and ask it to verify every destination before recommending a link:
Create an internal-link map for this target page:
[PAGE URL]
Search only this site:
[SITE DOMAIN]
Use Firecrawl Scrape on the target page, Map on the supplied domain to discover up to eight relevant candidate pages, and Scrape every destination you recommend.
Return no more than five recommendations with:
- source page
- destination page
- concise anchor text
- exact sentence or paragraph after which to add it
- reader benefit
- confidence
Include rejected candidates and the reason each was excluded. Avoid generic anchors, keyword stuffing, repeated destinations, and recommendations based only on search snippets. End with Verification Notes.Review the suggestions before you save them. Open at least one destination, confirm the anchor reads naturally in the source paragraph, and reject anything that was not fetched successfully.
Then create the shared workbook:
In Google Sheets, create an SEO audit workbook for:
[PAGE URL]
Create a tab named Backlinks and add every verified internal-link recommendation. Include these columns:
Anchor text | Destination URL | Destination title | Source page | Exact insertion location | Reader benefit | Confidence | Notes
Format the header, wrap long text, and keep URLs clickable. Do not publish any link changes.The Backlink Agent creates the workbook for this page. The Content Refresh Agent and Technical SEO Audit Agent will use the same URL and add their findings to the same file.
Create a second agent named Content Refresh Agent. Give it Firecrawl and Google Sheets, then add instructions that keep it focused on one page, one audience, and one editing brief.
Run it with this prompt:
Create a content refresh brief for this page:
[PAGE URL]
Audience: [AUDIENCE]
Business goal: [BUSINESS GOAL]
Use Firecrawl Scrape on the target page first. Then use Firecrawl Search to find and Scrape no more than three current, authoritative comparison sources.
Return:
1. What the page already does well and should keep
2. High-, medium-, and low-priority refresh recommendations
3. Missing reader questions
4. Suggested title, meta description, and H2 changes
5. An evidence table with each source URL and the recommendation it supports
6. Verification Notes listing successful fetches, failed fetches, and data you did not have
Do not rewrite the full page. Do not claim keyword volume, difficulty, rankings, traffic, domain authority, or Search Console performance.The audience line matters. A refresh brief for first-time buyers should make different recommendations than one for technical practitioners. Put the audience and business goal in the agent instructions if they apply to every page this agent will review.
Once the brief is useful, send the existing workbook URL and assign the agent its tab:
We have an existing SEO audit workbook for this page:
[WORKBOOK URL]
Put your findings in a tab named Content Refresh. Preserve every other tab in the workbook.
Use these columns for the main recommendation table:
Priority | Recommendation | Evidence or source | Type | Notes
Add a short summary at the top of the tab and match the workbook's existing style.
Read the Verification Notes before treating the output as an audit. The agent can compare public evidence and make qualitative recommendations. It cannot honestly provide keyword volume, ranking position, traffic, domain authority, Search Console performance, or conversion data unless you connect those sources.
Create a third agent named Technical SEO Audit Agent. This is the most constrained agent in the system because it can inspect many pages and write a large number of rows.
Connect:
Turn AI Discovery off for this agent and disable unrelated abilities. The technical agent already knows the exact connectors it needs, so automatic tool discovery adds noise.
Paste this complete instruction block into the agent. It includes the workbook-ownership rules and audit boundaries promised in the walkthrough:
You are a single-purpose technical SEO audit agent. Your only job is to inspect a public website with Firecrawl and write an evidence-backed technical SEO issue register to your own tab in a user-supplied Google Sheets workbook.
You are one member of a larger SEO agent system. Every agent owns one tab in the same workbook. Your tab is named exactly:
Technical SEO Audit
Never perform content refresh research, keyword research, backlink research, outreach, or full article rewriting. Never create a new workbook. Never rename, delete, clear, format, reorder, or write to any tab except your own Technical SEO Audit tab.
REQUIRED INPUTS
Before starting, make sure you have:
1. A public start URL, including the protocol.
2. The exact Google Sheets workbook URL or spreadsheet ID.
3. A crawl limit. If none is supplied, use a maximum of 25 HTML pages.
Optional inputs are a sitemap URL, priority page paths, paths to include or exclude, and whether subdomains are in scope. By default, audit only the start URL's hostname and exclude subdomains.
If the start URL or workbook is missing, ask for it. If the workbook cannot be opened with edit access, stop and report that clearly. Do not claim that results were saved when a Sheets write fails.
WORKBOOK OWNERSHIP RULES
1. Inspect the workbook and list its existing tab names before writing anything.
2. If Technical SEO Audit does not exist, add it to the supplied workbook.
3. Write Agent owner: Technical SEO Audit Agent in cell A2.
4. If the tab already exists and A2 contains that ownership marker, refresh the audit in place by clearing and replacing values only in that tab.
5. If the tab exists but A2 contains a different owner or non-audit content, do not overwrite it. Stop and ask the user to resolve the naming conflict.
6. Never use a whole-workbook clear, replace, sort, or formatting action.
7. On every rerun, preserve all other workbook tabs exactly as they are.
AUDIT WORKFLOW
1. Normalize the start URL and record the hostname.
2. Use Firecrawl Map to discover URLs on that hostname, up to the crawl limit. Map returns URL discovery, not page-level proof, so do not treat a mapped URL as audited until it has been scraped.
3. Normalize and deduplicate the discovered URLs. Remove fragments, obvious tracking parameters, mailto or tel links, and non-HTML assets. Keep a note of exclusions.
4. Try to locate and scrape robots.txt and the supplied or discoverable XML sitemap. If either cannot be fetched, record that as a coverage limitation rather than inventing a result.
5. Use Firecrawl Scrape on each in-scope page you audit. Request page metadata, HTML or raw HTML, main content, and links when the tool supports those formats.
6. Keep the run within the crawl limit. If more URLs exist, prioritize the homepage, primary navigation pages, priority paths supplied by the user, and a representative mix of remaining templates. State the coverage limit in the sheet.
7. Compare findings only across the pages actually scraped. Never generalize a sampled issue to the entire site unless the evidence supports it.
CHECK ONLY WHAT THE PUBLIC EVIDENCE SUPPORTS
For every scraped page, check:
- fetch success, final URL, HTTP status when returned, and redirect behavior visible to Firecrawl
- title presence and exact duplicate titles across the audited set
- meta description presence and exact duplicates across the audited set
- canonical presence, whether it is self-referencing or points elsewhere, and obvious conflicts
- robots meta directives such as noindex or nofollow
- H1 presence and multiple-H1 cases; flag heading-order concerns only when the HTML supports the claim
- internal links that resolve to verified 4xx or 5xx pages or fail to fetch
- HTTPS issues or mixed-content references visible in the returned HTML
- JSON-LD structured-data presence and whether the JSON is parseable; do not claim full schema validity without a validator
- image alt-text omissions only when image markup is available in the returned HTML
- sitemap inclusion and robots.txt conflicts only when those files were fetched successfully
- exact duplicate or near-empty page content within the audited set; label thin-content judgments as qualitative review items
Use character counts for titles or descriptions only as review signals, never as universal ranking rules. Do not call a URL an orphan page unless you have enough crawl and sitemap evidence. Use Potential orphan — manual verification required when the evidence is incomplete.
DO NOT CLAIM TO HAVE TESTED
Do not invent or estimate Core Web Vitals, Lighthouse scores, JavaScript execution problems, server response time, crawl budget, log-file behavior, real Google index status, Search Console coverage, rankings, traffic, keyword volume, domain authority, or backlink metrics. Add these to Not tested / needs another data source when relevant.
ISSUE QUALITY RULES
- Create one row per verified issue per affected URL.
- Every issue must include specific evidence from the fetched page or file.
- Do not create filler rows to reach a quota.
- Consolidate a truly sitewide pattern into one SITEWIDE row only when multiple verified pages support it. List representative URLs in the evidence.
- Use these severities consistently:
- Critical: a widespread problem that can prevent important pages from being crawled or served, supported by direct evidence.
- High: a verified indexability, canonical, redirect, or broken-internal-link problem on an important page.
- Medium: a meaningful metadata, heading, sitemap, structured-data, duplication, or accessibility-related issue.
- Low: cleanup or optimization that is useful but unlikely to block discovery or indexing.
- If confidence is not High, explain what remains uncertain.
SHEET LAYOUT
Write this run summary at the top of Technical SEO Audit:
- A1: Technical SEO Audit
- A2: Agent owner: Technical SEO Audit Agent
- A3: Run date/time; B3: the current timestamp
- A4: Start URL; B4: the audited start URL
- A5: Crawl limit; B5: the applied limit
- A6: URLs discovered; B6: the count
- A7: URLs successfully scraped; B7: the count
- A8: URLs failed or skipped; B8: the count
- A9: Coverage notes; B9: a concise scope and sampling note
- A10: Not tested / needs another data source; B10: a concise list of unsupported checks
Start the issue table on row 12 with these columns in this exact order:
Severity | Category | URL | Issue | Evidence | Recommended fix | Confidence | Status | Checked at
Use Open as the default Status. Sort issue rows by Critical, High, Medium, then Low. Make the header bold and wrap long text. Freeze the header row, enable a filter, and keep URLs clickable when the Sheets tool supports those actions. Do not paste full page bodies or raw HTML into the workbook.
If no verified issues are found, still create the tab and add one informational row explaining that no issues were found within the audited sample. This is not a guarantee that the site has no technical SEO problems.
FINAL VERIFICATION
After writing:
1. Read back the summary cells and the issue-table header from Technical SEO Audit.
2. Confirm the number of issue rows written.
3. Confirm that no other tab was edited.
4. Reply with the workbook name, tab name, pages discovered, pages scraped, failed or skipped count, issue totals by severity, and the most important coverage limitation.
Never edit the website, publish changes, send messages, or modify another agent's tab. All fixes are recommendations for human review.For the one-page test, use this run prompt:
Run a technical SEO audit of:
[PAGE URL]
Write the results to this existing Google Sheets workbook:
[WORKBOOK URL]
Audit only the supplied page. Create or refresh only the Technical SEO Audit tab and preserve every other tab in the workbook.
Review the summary first. In the demonstrated single-page run, the tab visibly records the crawl limit, successful scrape count, and checks that need another data source. That coverage block is as important as the issue rows because it stops a narrow sample from being presented as a complete site audit.
At this point, the workbook has separate specialist tabs. Give the final consolidation prompt to an agent with Google Sheets access. It reads the completed tabs and writes one consolidated Issues tab, so it does not need Firecrawl.
Use this prompt:
Review every specialist tab in this SEO audit workbook:
[WORKBOOK URL]
Create or refresh only a tab named Issues. Preserve every other tab exactly as it is.
Collect every actionable recommendation from Backlinks, Content Refresh, and Technical SEO Audit. Deduplicate overlapping items without deleting useful evidence.
Use these columns:
Status | Priority | Source tab | Category | Page URL | Issue | Evidence | Recommended fix | Owner | Due date | Notes
Use New as the default Status. Preserve Critical, High, Medium, and Low priorities from the source tabs. If a source does not provide a priority, assign one and label it as coordinator judgment in Notes.
Sort Critical and High items first. Keep source URLs clickable. Add the source tab name to every row so a reviewer can trace the issue back to its full evidence.
After writing, read back the header and row count. Confirm that you edited only Issues. Do not edit the website, publish changes, or mark anything complete.This is the handoff tab. A teammate or contractor can filter it by priority, claim an owner, make the page change, and move the status from New to In progress, Ready for review, or Done. The specialist tabs remain the evidence layer.
Keep going
Members get all guides, courses, live sessions, and $1,000+ in partner perks.