What should a technical SEO audit from an agency actually include?
A buyer guide to technical SEO audit deliverables: evidence, affected URLs, prioritization, implementation-ready tickets, acceptance tests, and measurement.
A technical SEO audit should reduce uncertainty enough for somebody to make and verify a change. That is a higher bar than exporting crawler warnings, assigning a health score, or presenting a long deck of observations. The buyer needs to know what population was inspected, what was actually observed, why the issue matters, which URLs or templates are affected, what should change, who should own it, and how to tell whether the implementation worked.
Google’s own documentation treats crawl access, HTTP responses, indexability, canonicalization, structured data, internal discovery, and search performance as separate systems. A useful audit preserves those distinctions. It should be able to say that a suspected technical problem is not supported by the evidence, or that a healthy technical layer should be left alone while the team works on page ownership, proof, local visibility, or conversion instead.
This guide is for a buyer comparing agencies or reviewing a proposed audit scope. It is not a list of every possible SEO check. It is a specification for the evidence and handoff you should expect to own when the engagement ends.
Ask for eight concrete deliverables, not an “SEO score”
| Deliverable | What it should contain | What it lets you decide |
|---|---|---|
| Scope manifest | Exact domains or subdomains, environments, URL populations, templates, priority services or locations, data sources, and exclusions. | Whether the audit actually covered the business surfaces you paid to inspect. |
| Crawl and indexation evidence | HTTP status, robots directives, noindex state, sitemap presence, internal discovery, canonical signals, and URL Inspection or Search Console evidence where available. | Whether an important page can be discovered, crawled, and considered for indexing. |
| Rendering and template evidence | Server HTML and rendered output where relevant, mobile behavior, metadata, links, structured data, and repeated template patterns. | Whether the problem is systemic, route-specific, JavaScript-dependent, or merely cosmetic. |
| Query-to-page ownership map | Priority commercial intents mapped to the canonical page that should satisfy each one, plus missing-page or cannibalization decisions. | Whether the site has a page architecture problem rather than a purely technical problem. |
| Finding ledger | Observed state, expected state, affected scope, evidence source and date, confidence, reproduction steps, and commercial/search impact. | Whether each recommendation is reproducible and worth attention. |
| Prioritized backlog | Recommended action ranked by commercial relevance, confidence, effort, dependencies, risk, and owner. | What should ship first and what can wait. |
| Acceptance tests | A pass/fail test for each implementation: source, rendered page, headers, crawl result, schema validation, or another observable condition. | Whether the fix was implemented correctly without waiting weeks for rankings. |
| Measurement plan | Immediate technical checks separated from later indexation, impressions, query fit, clicks, qualified inquiries, and revenue evidence. | Which outcomes can be judged now and which need a maturation window. |
The audit should begin with a scope manifest
You cannot judge completeness without knowing what population was tested. “We crawled the site” is too vague when the business may have a marketing domain, application subdomain, location pages, faceted URLs, JavaScript routes, legacy hosts, staging environments, and pages that only appear in Search Console or XML sitemaps. The handoff should identify the exact surfaces included and excluded.
A useful scope manifest records
- Canonical production hostnames and any relevant subdomains.
- Important page types and templates: services, locations, products, articles, tools, listings, or other commercial surfaces.
- Known URL populations from internal links, XML sitemaps, analytics, Search Console, logs, databases, or CMS exports when available.
- Priority services, locations, buyer journeys, and conversion paths that should receive greater commercial weight.
- Available Search Console, analytics, log, CMS, CDN, or server evidence and the date each source was read.
- Explicit exclusions and access limitations so a missing data source is not quietly treated as a clean bill of health.
This makes later claims auditable. If the agency says “all priority service pages are indexable,” the buyer can see which pages counted as priority and which evidence supported the statement.
Every material finding should carry its evidence with it
The evidence required depends on the claim. Google Search Essentials describes baseline technical requirements such as crawler access, successful HTTP responses, and indexable content. Google’s crawling and indexing documentation separately covers discovery, sitemaps, canonicalization, redirects, JavaScript, robots controls, and other mechanisms. The audit should point to the observed state of the relevant mechanism instead of substituting a generic label such as “indexation issue.”
| Finding | Evidence to expect | Weak substitute |
|---|---|---|
| Canonical conflict | Affected URL set, declared canonical, redirects, internal links, sitemap signals, and Google-selected canonical when available. | “Duplicate content detected.” |
| Indexation problem | Search Console Page Indexing or URL Inspection evidence, page response, directives, canonical state, discovery path, and representative URLs. | “Google does not like these pages.” |
| Rendering problem | Raw/server response compared with rendered output, affected elements, reproduction steps, and the template population. | A screenshot of one broken page. |
| Internal discovery problem | Links into the page population, anchor/context examples, orphan checks, sitemap state, and crawl depth or graph evidence. | “Add more internal links.” |
| Structured-data defect | Visible page claim, emitted JSON-LD or microdata, validation result, affected templates, and the relevant Google documentation. | A schema score without the markup or page context. |
| Performance concern | Field or lab source, metric, population, device context, date, and the suspected implementation cause. | A single Lighthouse number treated as universal field performance. |
A developer should not have to rediscover the problem
An implementation-ready recommendation is specific enough to turn into a ticket without repeating the investigation. The audit does not need to write production code for every issue, but it should define the affected rule, current behavior, expected behavior, dependencies, examples, risk boundary, and acceptance test.
| Ticket field | Example of useful detail |
|---|---|
| Affected scope | All /services/:slug pages generated by the service template; 42 URLs in attached population. |
| Observed state | Canonical points to the parent category rather than the route itself on the listed URLs. |
| Expected state | Each unique service route should self-canonicalize unless a documented consolidation rule applies. |
| Dependencies | Template helper and sitemap output use the same canonical URL function. |
| Risk / rollback | Changing canonical rules can alter consolidation signals; deploy to the bounded population and verify before expanding. |
| Acceptance test | Rendered head and HTTP response show the intended canonical on every sampled URL; sitemap emits the same URL; no unexpected redirect is introduced. |
Content and information-architecture recommendations need the same discipline. If the audit recommends a new service page, it should name the buyer intent, current competing or missing owner, evidence needed on the page, internal discovery path, and whether the proposal is a correction to an obvious gap or a lower-confidence experiment.
Follow one sample finding from evidence to a recheck
VOID’s representative audit register includes a Legacy URL row: a 301 reaches the priority URL, but internal links still use the old address. This is illustrative, not a client crawl or a completed fix. Here is the handoff a buyer should request.
What makes the sample finding actionable
- Evidence and scope: attach the source-page list, anchor text, old destination, status chain and observation time. A sample row does not establish how many links or templates are affected.
- Owner and decision: CMS/content owner — to confirm. Identify whether a template or individual pages own the links, then point those links directly to the intended destination. Keep the valid permanent redirect for older visitors and external links.
- Acceptance and status: re-crawl the listed source pages and verify their changed links reach the intended 200 destination directly. Keep validation Not started until the actual recheck; use Unknown when access or evidence is missing.
- Search outcome: record later indexed and performance observations separately. Closing this link repair does not establish an increase in rankings, clicks or inquiries.
Use browser check C21 to record Needs attention and a note with scope, owner, action and recheck. Its CSV export contains check_id, category, check, status and evidence_note; ownership stays in the note. The separate T21 evidence-template row has dedicated owner, confidence, recommended_fix, validation_method and validation_status columns. The two template downloads are blank starters; workspace exports contain the statuses and notes you enter.
Prioritization should follow commercial value and confidence
Severity labels inherited from audit software are not enough. A technically imperfect URL with no meaningful buyer role can be less important than one canonical defect on a page already earning high-intent impressions. A useful backlog combines technical severity with the business surface affected, confidence in the diagnosis, implementation cost, dependencies, and reversibility.
| Factor | Question the backlog should answer |
|---|---|
| Commercial relevance | Does the issue affect a priority service, location, buyer question, lead path, or valuable supporting page? |
| Reach | One URL, one template, one subdomain, or the entire indexable population? |
| Confidence | Is the defect directly observed, strongly inferred, or still experimental? |
| Effort | Can it be repaired in one rule or does it require redesign, migration, CMS work, or cross-team dependencies? |
| Risk | Could the proposed fix damage healthy routes, remove useful content, change canonical ownership, or break conversion paths? |
| Measurement | Can implementation be checked immediately, and what later search or commercial signal should be watched? |
The backlog should also contain “do not change yet” decisions. A recently revised page waiting on indexation or enough impressions is not improved by being rewritten every week. Restraint is part of the audit output when the evidence is immature.
Separate release acceptance from ranking outcomes
A clean release-verification chain
- Implement the bounded change against the documented affected scope.
- Verify the exact source or configuration that is intended to ship.
- Run the relevant mechanical checks: tests, build, link/render checks, schema validation, or other route-specific QA.
- Deploy the exact merged tree or version through the normal release path.
- Inspect the rendered production page, response headers, canonical, robots directives, important links, and changed content.
- Record the release date and freeze the experiment long enough for indexation and performance evidence to mature.
- Evaluate search and commercial outcomes separately from whether the implementation itself was technically correct.
This protects both sides of the engagement. A technically correct fix can be accepted immediately without pretending that rankings or leads have already improved. A later performance miss can then be investigated as a search, intent, competition, proof, or conversion question rather than being confused with whether the code shipped.
Use this scorecard before you hire the agency
| Ask the agency | A strong answer should explain |
|---|---|
| Which exact pages, templates, domains, and markets are in scope? | The population, priority surfaces, data sources, exclusions, and access constraints. |
| What evidence accompanies a material finding? | Affected URLs, observed state, source/date, reproduction method, and expected state. |
| How do you distinguish confirmed defects from experiments? | Confidence labels and different release/evaluation rules for corrections versus hypotheses. |
| How do you prioritize the backlog? | Commercial relevance, reach, confidence, effort, dependencies, risk, and measurement—not crawler severity alone. |
| What does an implementation-ready ticket contain? | Scope, current behavior, expected behavior, examples, owner/dependencies, risk, and acceptance test. |
| How will we know a fix shipped correctly? | A mechanical or rendered production check that does not depend on waiting for rankings. |
| Which outcomes need a later observation window? | Indexation, impressions, CTR, query fit, qualified inquiries, and revenue where measurement permits. |
| What happens if the site is technically healthy? | The engagement can shrink, redirect to a better-supported problem, or end without manufacturing defects. |
| Do we own the evidence and backlog? | The buyer receives the files, populations, tickets, notes, and handoff needed to implement internally or with another vendor. |
Audit red flags
- A single health score is treated as the main conclusion.
- Hundreds of warnings are presented without affected populations or commercial priority.
- Recommendations do not name example URLs, templates, or expected behavior.
- The seller guarantees rankings, traffic, AI citations, leads, or revenue from technical fixes.
- A monthly retainer is required before the buyer can receive or use the audit findings.
- Healthy or recently changed pages are scheduled for rewrites without performance or defect evidence.
- There is no distinction between implementation acceptance and later search outcomes.
How VOID scopes the handoff
VOID’s focused diagnostic currently starts at $2,500 and is scoped for 5–10 business days. The purpose is to define the problem with enough evidence to support a build or a no-build decision. Implementation is separate, and a no-build handoff is available. Depending on scope, the work can cover crawl and indexation, rendering, page/query ownership, local and entity signals, proof gaps, contact-path measurement, and a ranked implementation backlog.
| Observed situation | Useful handoff |
|---|---|
| Confirmed technical defect blocks valuable pages | Implementation-ready tickets, bounded affected population, acceptance tests, and release priority. |
| Technical layer is healthy but intent ownership is weak | Query/page map, content or architecture decision, proof requirements, and measurement plan. |
| Evidence is incomplete | Explicit unknowns, missing access/data, next diagnostic step, and no fabricated certainty. |
| Recently shipped work is waiting on evidence | Freeze period and the exact future signal that would justify another change. |
| Nothing material is wrong | No-build handoff and a short list of conditions worth monitoring. |
Source ledger
These sources support the operating guidance above. Platform behavior and documentation can change, so volatile implementation details should be rechecked before a rollout.
- Google Search Essentials — Google Search Central. Baseline technical, spam, and content requirements used to distinguish eligibility and crawl/indexation facts from later performance outcomes.
- Crawling and indexing — Google Search Central. Google documentation covering crawling, sitemaps, robots controls, canonicalization, redirects, JavaScript, and related indexing systems.
- How to specify a canonical URL with rel="canonical" and other methods — Google Search Central. Primary documentation for canonical signals and consolidation methods referenced in the evidence examples.
- UrlInspectionResult — Google Search Console API. Documents URL Inspection result fields used as an example of URL-level Google index-status evidence.
- Understand how structured data works — Google Search Central. Primary guidance for structured-data eligibility and the relationship between markup and visible page content.