Agency technical SEO · audit to ticket
Turn crawl and Search Console exports into SEO tickets an engineer can verify
ZeroTwo reads the exports you upload and drafts the ticket. Each ticket keeps its evidence chain: the crawl row it came from, a second source that corroborates it, the fix, and the test that proves it shipped. A finding that rests on one export is labelled a hypothesis, not a fact.
Ticket structure · a template, not a client
Crawl row
Export file name, the row or URL, and the status, canonical and indexability values exactly as exported.
Corroboration
A second source that agrees: URL Inspection (indexed version and live test), rendered HTML, or a server response.
Proposed fix
The change, who owns it, the risk, and how to roll it back.
Acceptance test
The re-crawl or re-inspection result that proves the fix shipped.
Label every finding by how much evidence it has
An export shows a symptom. Whether it is an issue, a probable cause or a guess depends on what else you have checked, so the label travels with the ticket and a ticket only gets to say "fix" when it is Confirmed.
Confirmed issue
Two independent sources agree and someone has seen the failure.
Example: A URL the crawl reports as 404 that a live URL Inspection test and a direct request also report as not found, and that pages still link to.
Ticket it as a fix.
Probable cause
One explanation fits the export, but a second source has not ruled out the others.
Example: Pages missing from the index that are also orphaned in the crawl. Orphaning fits, but the stated indexing reason has not been checked.
Verify first, then promote or demote.
Unverified hypothesis
A pattern in a single export that could have several causes.
Example: Low click-through at positions 4 to 10. It could be the title, a SERP feature, branded queries, or normal for that query.
Write it as a test, not a fix.
What each export can and cannot tell you
Every row below is a signal agencies commonly turn into a ticket. The last column is the bar a ticket has to clear before it goes to a developer.
| Signal in the export | What it supports | What it does not prove | Corroborate with | Ticket-ready when |
|---|---|---|---|---|
| 4xx in the crawl | The URL returned an error to the crawler at crawl time. | That anything links to it, or that it matters for search. | Inbound internal links from the crawl, a live URL Inspection test, server logs if you have them. | It is linked from somewhere that matters and still fails a live check. |
| Soft 404 flag | The page looks empty or error-like under the crawler's rules. | That Google treats it as a soft 404. | The reason Search Console reports, and the rendered HTML. | Search Console lists it and the rendered HTML confirms thin or error content. |
| Not indexed in Search Console | Google's stated reason for the URL at its last check. | Why it really happened, or that the URL should be indexed at all. | URL Inspection (indexed version against live test), canonical and robots values in the rendered HTML, internal links. | The reason survives a live test and the page is meant to rank. |
| Positions 4 to 10 with low CTR | Queries where the page appears below the top three. | That a better title would earn more clicks. | Query-level data, the live results page for the right location and device, brand against non-brand split. | CTR is low against your own similar queries and you have looked at the results page. |
| Traffic drop | Fewer clicks or sessions in a period. | The cause: seasonality, tracking, rankings, demand and indexing all look alike. | Release and update dates, impressions against CTR, analytics annotations. | One cause is isolated. Until then it stays a hypothesis. |
| Core Web Vitals failing | A group of URLs misses a field threshold. | Which element or script is responsible. | Field data by metric, then a lab trace for one page per template. | A trace names the element or interaction and the fix has been tested. |
| Missing schema | Your crawl's extractor did not find a markup type. | That the page is eligible for it, or that markup is absent after rendering. | Rendered HTML, Search Console enhancement reports, Google's documentation for that type. | The page is eligible and the visible content matches the markup. |
| Underlinked pages | Few internal links point to the page in the crawl. | That adding links will lift rankings. | Crawl depth, site structure, whether the page is worth promoting. | The page matters commercially and a specific linking source is named. |
Anatomy of one source-linked ticket
This is the structure ZeroTwo drafts into, shown as a blank template. It is not a worked example from a client, so no figures appear in it.
The last field matters most. If the draft cannot fill a field from evidence you pasted, it should say so rather than invent a plausible answer.
Open the ticket prompt in ZeroTwoTicket template
No client data- Title
- One sentence naming the URL pattern and the defect.
- Evidence label
- Confirmed, Probable or Hypothesis.
- Source rows
- Export file name and the exact rows or URLs.
- Corroboration
- URL Inspection result, rendered HTML check, response header, or log line, each with the URL it was run on.
- Proposed fix
- The specific change, written for the owner who ships it.
- Owner
- Development, content, or both, named.
- Risk and rollback
- What could break and how to undo it.
- Acceptance test
- The re-crawl or live-test result that closes the ticket.
- Not yet verified
- Anything above that rests on assumption, listed plainly.
Positions 4 to 10: state the model before you quote a number
In Search Console, CTR is clicks divided by impressions, and the table value for average position is the average position of the topmost result for that row. Google also notes that results differ by time, place, device and history, so a query in your export may not reproduce for you.
Inputs you must state
- Where the CTR curve comes from: your own Search Console data for similar queries, with the date range stated.
- Whether branded queries are separated out. They distort any average.
- That average position is an average, so a page can appear at several positions across the period.
- Which results-page features appear for these queries and could take clicks.
- Seasonality across the date range you pulled.
What the output may claim
- A low, expected and high range per page, never one number.
- The assumptions written next to the range.
- A test to run (new title, new snippet) rather than a promised gain.
- No click or revenue forecast when the inputs above are missing.
Source: Google Search Console Help, Performance report overview, support.google.com/webmasters/answer/7576553, checked 5 October 2026.
Core Web Vitals: field data finds the page group, a trace finds the cause
A failing report row tells you a group of URLs misses a threshold. It does not tell you what to change, so no CWV ticket should prescribe a fix before a performance trace names the cause.
- Largest Contentful Paint, loading
- 2.5 s
- Interaction to Next Paint, interactivity
- 200 ms
- Cumulative Layout Shift, visual stability
- 0.1
"Good" thresholds, measured at the 75th percentile of page loads. Source: web.dev/articles/vitals, checked 5 October 2026.
Group by field data
Take the failing URLs from the field report and group them by template and by metric, not by page.
Record a lab trace
Trace one representative URL per template in DevTools. Lighthouse does not report INP, so use a trace for interaction problems.
Name the cause
The LCP element, the long task, or the element that shifts. An LCP failure can be server response or render-blocking resources, not only images.
Ticket with a target
State the metric and threshold, test the fix in the lab, then re-check field data later. Field data reflects real visits, so it moves after a release, not at it.
Schema findings need an eligibility check before a spec
A crawler reporting "no structured data" is a prompt to look, not a prompt to add markup to every page type. Template-wide recommendations skip two questions: is the page eligible, and does markup already exist once the page is rendered?
Before drafting a spec
- Confirm the page type and that Google documents rich results for that markup type.
- Check the rendered HTML, not just the source, for markup that already exists.
- Make sure the visible page content matches every property you would add.
- Write the spec per template, and list the pages that are not eligible.
How an audit session runs in ZeroTwo
Five stages, with the owner of each. ZeroTwo drafts at the stages where drafting helps. Verification stays with you.
Upload the exports
YouCrawl export, Search Console page and query exports, the Core Web Vitals report and any engineering notes. Upload only what the client agreement lets you share with a third-party tool.
Triage and label
ZeroTwo draftsZeroTwo groups rows by pattern and proposes an evidence label for each group, with the second source that would confirm it.
Corroborate
YouRun the URL Inspection tests, rendered-HTML checks and lab traces the triage asked for, and paste the results back in.
Draft the tickets
ZeroTwo draftsOne ticket per confirmed group in the format above, with unverified items listed as such.
Package the roadmap
You decideSequence tickets into sprints by owner and verification step. Impact estimates appear only when you supply the model behind them.
Prompts that ask for evidence, not conclusions
Each one names the export it expects and tells the model what not to guess. Open one in ZeroTwo and attach the file.
Crawl triage
From this crawl export, group the 4xx, 5xx, redirect-chain and non-indexable rows by pattern. For each group give the count, three example URLs, and which second source (URL Inspection, rendered HTML, logs) would confirm it. Do not recommend fixes yet.
Crawl_Export.csvOpen in ZeroTwoPosition 4 to 10 review
From this Search Console export, list pages with average position 4 to 10. For each, say what else could explain low CTR (branded queries, results-page features, query mix) and what I would need to check. Do not forecast clicks.
GSC_Pages_Queries.csvOpen in ZeroTwoIndexing reasons
From this page indexing export, group URLs by Google's stated reason. For each group, say which URLs I should test with URL Inspection first and what result would change the diagnosis.
Page_Indexing.csvOpen in ZeroTwoDraft one ticket
From the evidence below, draft one ticket with: evidence label (Confirmed, Probable or Hypothesis), source rows, corroboration, proposed fix, owner, risk and rollback, acceptance test. Write 'not yet verified' for anything without evidence.
Evidence_Notes.txtOpen in ZeroTwoCore Web Vitals grouping
From this Core Web Vitals export, group failing URLs by template and metric. For each group, list the lab trace I should record before any fix is proposed.
CWV_Report.csvOpen in ZeroTwoSchema eligibility
For these page types, list which schema types Google documents as eligible, what visible on-page content each requires, and how I should check the rendered markup. Do not recommend markup for pages I have not confirmed are eligible.
Page_Types.txtOpen in ZeroTwoRoadmap from labelled tickets
Turn these labelled tickets into a four-sprint roadmap: sprint goal, owner (development or content), tickets, and verification step. Leave traffic impact blank unless I give you a stated model.
Tickets.txtOpen in ZeroTwo
Choosing the tools around this workflow
The crawler, Search Console, field data and trace tools produce the evidence. ZeroTwo works on the exports from them. If you are still deciding which AI SEO products to use for the wider job, the AI SEO tools comparison covers product selection. ZeroTwo publishes both pages, so read them with that in mind.
Questions agencies ask before trusting the output
Can AI work out why pages are not indexed from an export alone?
Not reliably. A Search Console reason code plus a crawl row narrows the candidates, but a live URL Inspection test and the rendered HTML are what confirm a cause. In this workflow ZeroTwo groups the URLs and tells you which to inspect first. It does not replace the inspection.
Is a list of pages at positions 4 to 10 a click forecast?
No. Average position is an average of the topmost result across the period, and click-through depends on query mix, brand queries and what else appears on the results page. Report a range with the CTR source and date range beside it, or report a test to run instead of a number.
What should I upload?
Whatever your tools export: a crawl file, Search Console page and query exports, a page indexing export, a Core Web Vitals report, and any engineering notes. Upload only what your client agreement allows you to share with a third-party tool, and split very large exports by template or site section so each prompt asks one question.
Does ZeroTwo replace a crawler, Search Console or a log analyzer?
No. Those tools produce the evidence. ZeroTwo works on the exports you bring to it, drafts the grouping, the tickets and the roadmap, and leaves the verification steps to you.
What does ZeroTwo cost?
The Free plan is $0 with 100 free credits to start, 20 bonus credits every day you log in, and no card required. Plus is $14.99 per month and Pro is $29.99 per month ($26.99 per month billed annually). Check the pricing page for the limits on each plan.
More agency workflows in ZeroTwo
- Agency content briefs
Once decay and CTR tests are confirmed, turn them into briefs your writers can execute.
- Agency account audits
Technical SEO is one layer of an account audit that also covers paid, analytics and CRM.
- Agency keyword research
Tie technical findings back to the keyword clusters worth targeting.
Bring one real export and see what survives the evidence check
Upload only data you are allowed to share, and verify every ticket before it goes to a client.