Site audit & broken links

Audience: Merchant·Plan: free·Read: 9 min·Updated: 2026-08-03

In 30 seconds

• Scores every product and collection, tracks the score trend over time, and estimates revenue at risk from what is failing.
• AI fixes propose new titles and meta descriptions; you see a diff before anything is written, and every change is versioned.
• The broken-link manager pulls from four sources — catalogue scan, site crawl, real 404 hits from your storefront, and manual entry.
• Redirect targets are validated before they are written: a redirect to the broken URL itself, or to another dead page, is refused.
• Every applied redirect can be rolled back, and the source page's HTML can be rewritten so the dead link stops existing at all.

How to use it

1. Run a full-store audit. It scores every product and collection against the signal set.
2. Work the priority panels — they surface the highest-value failures rather than the first alphabetically.
3. Use AI fix on the ones you accept, reviewing the diff before applying.
4. Scan for broken links, then fix them in revenue order.
5. Schedule the audit so the score trend keeps building.

⏱ ~15 min first pass · 💳 Free (AI fixes and automation Starter+) · 🎯 Fix what costs money first

The audit

Every product and collection is scored across the signal set — title, meta description, headings, content depth, images and alt text, structured data, canonical and indexability, internal links, and URL structure. Each entity gets a grade, and the store gets a rolled-up score.

Surface	What it gives you
Score ring + trend	The store score today and its history, so improvement is visible
Category bars	Which signal families are dragging the score
Revenue at risk	What the failing entities are worth, so the list has a business order
Audit table	Every entity, filterable by grade and score, with inline SEO editing
Priority panels	The highest-impact failures, surfaced above the table
Detail drawer	Per-entity signal breakdown
Scheduled audits	Recurring runs on a cadence you set, with an emailed report

Inline editing and AI fixes

You can edit a title or meta description inline, directly in the table — the write goes to Shopify against the correct mutation for the entity type, so a collection edit does not run the product mutation and fail.

The AI fix flow proposes replacements, shows you a diff against the current value, and applies only what you accept. Fixes can be run in bulk as a tracked background job, with progress narrated as it goes.

Every SEO field change is written to revision history, so you can see what changed, when, and by whom — and those revisions feed the change timeline on Technical health when you are diagnosing a regression.

Broken links

A broken link is a URL on your site that returns an error. The manager tracks them from discovery through to fixed, with a status per row.

Four discovery sources

Source	What it finds	Signal quality
Catalogue scan	Dead handles across your products, collections and content	Potential breakage
Site crawl	Dead URLs found by following your real link graph, with the pages that link to them	Potential breakage
Live 404 beacon	Real 404 hits from visitors and Googlebot, fired from your storefront's not-found page	Actual breakage, happening now
Manual / import	URLs you know about	As you enter them

The live 404 beacon is the one that changes prioritisation. Every other source tells you what could be broken; the beacon tells you a dead URL is taking 500 hits a day right now. It requires the storefront app embed to be enabled.

Revenue at risk

Each row carries a calibrated revenue-at-risk figure computed from one shared model, so the number on the row, the number in the digest email and the number the risk register reports are always the same. Attributed revenue, organic clicks and real 404 hits all feed it.

That figure drives the priority grading:

Priority	Rule
Critical	5xx — the server is failing, not just the URL
High	4xx that demonstrably costs something — revenue, organic clicks, or real hits
Medium	Everything else

Fixing

For each broken URL, the suggester finds the closest still-live candidate using handle similarity, token overlap and title overlap, and grades its confidence.

Redirect targets are validated before they are written

Creating a Shopify redirect to whatever target was typed produced two failure modes that are painful to unpick: a redirect from a URL to itself (an instant loop), and a redirect to another dead page (a silent broken target that only shows up later as a chain problem).

Preflight validates the target before the redirect is written. Applied redirects are then annotated with chain health, so a redirect that has since become part of a chain or landed on a dead page is visible rather than assumed fine.

Three fix paths:

Path	What it does
Apply redirect	Creates a validated Shopify URL redirect
Fix the source page	Rewrites the dead URL inside the linking page's body HTML, so the broken link stops existing
Ignore	Records the decision without fixing

Rollback undoes a previously applied redirect using the recorded audit entry, so a wrong fix is reversible rather than permanent.

Supervised auto-fix

Auto-fix is opt-in and deliberately narrow. When enabled, it applies only the safest subset of stored suggestions without a human: a genuine match — never a tiered fallback — at or above the confidence bar, and at or below the money guardrail.

The reasoning is that a triage backlog never empties if every fix needs a human, but an automated redirect on a high-value URL with a low-confidence match is exactly the change you want a person to look at.

Operational view

Metric	What it tells you
MTTR	How long a broken link survives before it is fixed
Backlog age burndown	Whether the backlog is shrinking or just being added to
High-value 404s past SLA	The rows that are bleeding and overdue

A weekly digest emails the same picture.

Plan tiers

Capability	Free	Starter	Pro	Agency	Enterprise
Full-store audit + scoring	✓	✓	✓	✓	✓
Score trend + category bars	✓	✓	✓	✓	✓
Inline SEO editing	✓	✓	✓	✓	✓
Broken-link scan + manual fixes	✓	✓	✓	✓	✓
Redirect preflight + rollback	✓	✓	✓	✓	✓
Revenue at risk	—	✓	✓	✓	✓
Live 404 beacon	—	✓	✓	✓	✓
Redirect suggestions	—	✓	✓	✓	✓
Source-page rewrite	—	✓	✓	✓	✓
SLA, MTTR and burndown	—	✓	✓	✓	✓
AI fixes (single + bulk)	—	✓	✓	✓	✓
Scheduled audits + emailed report	—	✓	✓	✓	✓
Supervised auto-fix	—	—	✓	✓	✓

Revenue at risk depends on Search Console and attribution data.

Best practices

• Enable the 404 beacon. It is the only source that tells you what is breaking now rather than what might be.
• Fix in revenue order, not in scan order. The list is already sorted for you; overriding that is how a week gets spent on URLs nobody visits.
• Prefer fixing the source page over adding a redirect. A redirect is a permanent tax on every request; removing the dead link is free afterwards.
• Review the AI diff. The proposals are good, not infallible, and the diff takes five seconds to read.
• Schedule the audit weekly. The score trend is the artefact that makes SEO work visible to everyone else.
• Turn on auto-fix only after two manual passes. You want to see the suggester's confidence calibration on your own catalogue first.

See also

• Site crawler — one of the four discovery sources, and the remediation workbench
• Technical health — indexability and regressions
• Risk register — where audit and link findings are ranked together
• Opportunities — the revenue-side worklist