On Belay

Website Optimization Agent

Your site has problems you can't see. Dead clicks. Leaky funnels. Pages losing rank.

The Website Optimization Agent reads your analytics and search signals every week, diagnoses what's holding the site back, and hands you a reviewed list of recommendations — each one ready to approve, dismiss, or route to the fieldset that can act on it. It diagnoses; you decide; other fieldsets execute.

What it does

The Website Optimization Agent is a recommendation engine. Every week it pulls from your connected data sources, runs a set of specialist sub-agents over that data — site audit, SEO, AI visibility, traffic, and funnels — and produces concrete optimization recommendations.

Crucially, it doesn't change your site. It recommends. When you approve a recommendation, the agent routes the work to the fieldset that can carry it out — a meta or content fix goes to the AI Visibility Engine, a product-description change goes to Merchandising — or, if no installed fieldset can act on it, it stays in a "For Human Review" lane for your team. This separation is deliberate: the agent is the brain, your other fieldsets and your team are the hands.

Who it's for

Shopify brands that want a steady, weekly read on what's hurting conversion and search performance — without standing up an analytics practice. If you have the data flowing in but not the time to mine it for action, this fieldset turns it into a short list of things worth doing.

What the cycle needs (data sources)

Enabling the fieldset doesn't gate on any one integration — you can switch it on first and connect sources as you go. What matters is what's connected when a cycle runs: the agent runs a health probe before every cycle and uses whatever you've connected, so each weekly cycle only runs the diagnostics whose required sources are connected and healthy. The more sources you connect, the more of its specialist analyses produce findings; the cycle degrades gracefully around anything that's missing.

The agent can pull from six sources:

  • Google Analytics 4 — sessions, conversions, traffic source. Powers the traffic and funnel analyses.
  • Shopify — orders, revenue, product sales. Powers the site-audit and funnel analyses.
  • Google Search Console — query impressions, CTR, rank. Powers the SEO analysis; without it the SEO recommendations sit the cycle out.
  • Microsoft Clarity — session replays, dead clicks, rage clicks. Powers the behavioral signals behind the funnel findings; without it those signals sit out and the cycle still runs.
  • Ahrefs — competitor outranking and keyword gaps. Adds competitive context.
  • DataForSEO — SERP context and the broader keyword universe. Adds competitive context.

Connect everything at Dashboard → Integrations. GA4 and Shopify carry the core traffic, funnel, and site-audit findings, so connect those to get the most out of each cycle; add Search Console for SEO recommendations, Clarity for behavioral and UX signals, and Ahrefs/DataForSEO for competitive context.

Setup (step-by-step)

  1. Enable the fieldset. Go to Dashboard → Fieldsets, find Website Optimization Agent, and enable it (admin only). You'll pass through the consent step. Enabling doesn't gate on any integration — connect your sources next.
  2. Connect GA4 and Shopify at Dashboard → Integrations. These carry the core traffic, funnel, and site-audit findings, so a cycle produces the most when both are connected and healthy. The agent checks each source's health at run time, not at enable time.
  3. Add the sources that produce more findings. Connect Google Search Console to turn on the SEO analysis, Microsoft Clarity for behavioral and UX signals (dead clicks, rage clicks, session friction), and Ahrefs / DataForSEO for competition-aware context. The agent uses whatever's connected and degrades gracefully around what isn't.
  4. Configure the fieldset in its Settings: set your goals, freshness thresholds, the Slack channel for the weekly digest, and any off-limits products.

Configuration

Everything is on the Website Optimization Agent's own pages:

  • Source Health Panel — persistent across the fieldset, it shows at a glance whether each data source is healthy and fresh, so you know the cycle has good inputs.
  • Cycle Status Banner — shows where the current weekly cycle stands.
  • Settings — goals, sources, freshness thresholds, Slack channel, and cadence.

How it runs (schedule)

The cycle runs weekly — Mondays at 8am UTC. Each cycle probes the sources, fetches data, runs the sub-agents over their slices, synthesizes findings into recommendations, and posts a digest to Slack. The probe decides what runs: each diagnostic only runs when the sources it needs are connected and healthy, so the cycle pauses or degrades a diagnostic whose required source is missing rather than producing weak recommendations — and the Source Health Panel tells you why. If Search Console is missing the SEO analysis sits out, if Clarity is missing the behavioral signals sit out, and if Ahrefs or DataForSEO is unavailable the relevant sub-agent runs without the extra competitive context — the rest of the cycle still completes.

Review & approval

The front door is the Recommendation Review queue — not a dashboard of charts, but the list of work that needs your judgment. Recommendations are grouped by where they can go: those that route to Merchandising, those that route to the AI Visibility Engine, and a "For Human Review" group for anything no installed fieldset can act on.

For each recommendation you can:

  • Approve / route it to the sister fieldset that will execute it (with a context-aware action label).
  • Edit before sending — your edits are captured to sharpen future recommendations.
  • Reject it, choosing from curated reasons plus free text.

Bulk actions exist for sections with several items, but bulk approve never dispatches silently — it opens a confirmation that names every recommendation being approved. No work item is dispatched without an explicit human click.

Outputs & where they land

  • Recommendations live in the Recommendation Review queue inside the fieldset.
  • Approved work is routed to the executing fieldset (Merchandising or AI Visibility Engine), where it flows through that fieldset's own review and approval before anything reaches your live store. The Website Optimization Agent itself never writes to Shopify.
  • A weekly digest is delivered to the Slack channel you configure.

Billing

The Website Optimization Agent is included in your On Belay plan — there's no per-fieldset, per-run, or usage charge, and the fieldsets it routes work to (Merchandising, AI Visibility Engine) are included too, with no per-publish cost. The only thing you pay for is your base plan tier, billed through Shopify Managed Pricing on top of your 14-day free trial. See Plans & billing for tier details.

Tips

  • Get GA4 and Shopify green first. They carry the core traffic, funnel, and site-audit findings, so getting them healthy gets the most out of each cycle. The Source Health Panel tells you exactly what's connected and fresh.
  • Add Search Console to unlock SEO. The SEO recommendations only run when GSC is connected, so add it as soon as you can.
  • Add Clarity for the behavioral picture. Dead clicks, rage clicks, and session friction sharpen the funnel findings; without it the cycle still runs, just with less UX signal.
  • Add Ahrefs and DataForSEO for competitive context. They turn good recommendations into competition-aware ones.
  • Use Edit before sending. Your edits both improve the dispatched work and teach the agent what good looks like for your brand.
  • Treat the queue as a weekly ritual. The cycle runs every Monday; a short review each week keeps the highest-value fixes moving.

Related