For AI agents: a documentation index is available at /llms.txt
How we work
How WorkPods works
At WorkPods, an AXO (agent experience optimization) and GEO agency with technical SEO underneath, every engagement follows four steps: Audit, Fix, You approve, Measure. AI agents run the repetitive checks and first drafts, operators decide what matters and edit every deliverable, and your team approves each change in our approval log before release.
Add /llms.txt and fix the demo form labels
- Prepared by
- Agent
- Checked by
- Operator
- Evidence
- Agent task blocked at step 3
Every approval is recorded in the approval log.
01
Split the work across agents, operators and your team
Each step has a job for our AI agents, a job for our senior operators and a job for your team. Agents handle the volume, operators make the judgement calls, and you keep the final say, so unedited drafts and risky patches stay off your live site.
| Step in the loop | What our AI agents do | What our senior operators do | What your team does |
|---|---|---|---|
| 1. Audit | Run buyer prompts across six AI surfaces, crawl HTML and structured data (labels telling machines what a page is), check directories and parse logs | Verify every finding against primary sources, cut false positives and rank fixes by pipeline impact | Confirm 3 to 5 key money-page tasks and review baseline evidence across GEO, AXO and SEO |
| 2. Fix | Generate initial file templates, schema patches and source-gap research notes | Write and edit every final page update, llms.txt (a short file telling AI tools what your site covers), robots.txt rule, JSON-LD patch and WebMCP form markup (an emerging standard) | Answer product or policy questions where internal business context is needed |
| 3. You approve | Package each deliverable with its evidence and target URLs | Sign off that the deliverable meets our editorial standards and technical checks | Review the exact text or patch under role-based access and record sign-off in the approval log |
| 4. Measure | Re-run buyer questions across AI engines, re-crawl updated HTML and test key tasks with a safe browser agent | Compare results before and after each fix across all defined measures and plan the next batch | Review the before-and-after proof and confirm priorities for the next cycle |
02
Approve every file and page fix before release
Your team approves the exact wording and patch for every deliverable, in our role-based approval log, before anything ships. You see each draft in full, and the log keeps a record your marketing, web and security teams can check later.
Deliverables fall into four groups.
- GEO content and page fixes: rewritten answer sections, vendor comparison pages and company descriptions with every factual claim tied to a named source.
- Readable website files: a draft llms.txt file, Markdown pages, Content-Signal and AI-crawler rules in robots.txt, JSON-LD patches, and HTML updates without JavaScript.
- Operable form updates: clear form field label fixes and WebMCP tool markup for key forms such as demo requests and contact forms.
- Callable platform specs: directory checks and tool description reviews with safety hints (like readOnlyHint and destructiveHint). We also scope one remote MCP server with OAuth 2.1 when needed.
Client data stays in WorkPods' own infrastructure; trust and data handling covers the details.
03
Measure every change before and after each fix
We measure every approved batch before and after the fix, across AI answers, site files, tasks, directories and server logs. We never guarantee rankings, citations, app picks or traffic, so the reports stick to defined measures you can check yourself:
- Named means your brand appears in the AI answer, Cited means your page is a source, and Cited instead names the sources used instead.
- Each AXO check is scored Pass, Partial, Fail or Not applicable. A missing optional file like llms.txt is Not applicable, never a fail, and only a present, broken file gets Fail.
- For Operable tasks, we test 3 to 5 key tasks from your money pages confirmed by your team. Each task ends Completed, Reached the confirmation step, Partial or Blocked, recorded with the step where it stopped.
- For Callable platforms, we check the official MCP Registry, Claude connectors, ChatGPT apps and Google connected apps. We mark your brand and up to five rivals as Listed or Not listed in each directory.
- For Observed agent visits, each log entry is Verified inside published IP ranges, Mismatch outside them, or Unverified without published ranges.
Search health is tracked in your Google Search Console and Google Analytics data alongside these measures. The first loop starts with a free GEO and AXO audit.
Questions
Questions buyers ask
Can WorkPods agents push changes to our website without our permission?
No. Our AI agents never push changes to your live website or CMS. They run the repetitive checks and prepare first drafts inside our workspace. An operator reviews and edits every file and page fix, and nothing ships until an authorised person on your team approves it in our approval log.
How does the approval log work for our marketing and web teams?
The approval log records every proposed deliverable, the operator who checked it, and the member of your team who approved it. Access is role-based, so you control who has view-only access and who can sign off on content edits, robots.txt policies, JSON-LD patches or form updates.
What happens when your browser agent tests our demo or checkout forms?
It walks through the task and stops before submitting anything: our browser agent never submits a form or buys anything. It tests 3 to 5 key tasks proposed from your money pages and confirmed by your team, such as finding pricing or booking a demo. Each task ends Completed, Reached the confirmation step, Partial or Blocked, with the step where it stopped.
Who implements the approved fixes on our site?
Your web, marketing operations or engineering team applies the approved files and patches in your CMS or codebase using the exact drafts we hand over. We hand over ready-to-use llms.txt files, robots.txt rules, JSON-LD patches, label fixes and edited copy, so your team does not have to write specs from scratch.
How do you measure whether the approved changes worked?
We retest before and after each fix across all defined measures. We check whether your pages and files move to Pass and whether tasks reach Completed or Reached the confirmation step. We also re-run buyer prompts for Named, Cited and Cited instead across six AI engines and check Google Search Console.
How can we evaluate your operators before signing an engagement?
Request the free audit first and judge the work before you commit. The form asks only for your work email and website. An operator checks your AI citations, agent readiness and technical SEO basics, and delivers the report within 3 working days.