EXTRACTIVE SUMMARY
A semantic SEO audit judges a property against its own topical map rather than against a checklist. Most audits return a list of forty findings instead of a verdict, because listing is mechanical and prioritizing is not. An audit collects five evidence sets: coverage against the tuple inventory, extraction quality across a heading sample, the internal link graph, information gain across core pages, and publication history. The evidence separates into symptoms and causes, and eight common symptoms trace to a smaller set of causes that no checklist distinguishes between. Findings that conflict are resolved by five precedence rules: technical precedes everything, a cause is never remediated after its symptom, coverage precedes extraction, prominence outranks volume at equal precedence, and a finding recurring across three clusters is systemic rather than local. The verdict names one binding constraint. An audit cannot tell you demand, off-site standing, or whether the subject was worth choosing.
What a semantic SEO audit judges
A semantic SEO audit judges a property against its own topical map, measuring what the property covers against what complete coverage of its subject would require.
This is what distinguishes it from a technical SEO audit and from a content audit. A technical audit judges a property against a standard of correct implementation. A content audit judges pages against a standard of quality. A semantic SEO audit judges a property against a denominator, and without a topical map supplying that denominator the audit has nothing to judge against and reverts to a checklist.
The first output of an audit on a property with no topical map is therefore the map itself. An audit that skips this step and proceeds to findings is evaluating pages against the auditor’s intuition about the subject.
Why most audits return a list instead of a verdict
Most audits return a list of findings rather than a verdict, because collecting findings is mechanical and prioritizing them is judgment.
A crawl returns hundreds of observations within an hour. Every one of them can be written up, severity-tagged, and formatted, and the resulting document looks thorough in direct proportion to how little decision-making it contains. A forty-item findings list transfers the entire prioritization problem to the client, who commissioned the audit precisely because they could not prioritize.
Two consequences follow. The client works the list from the top, and list order is usually severity-by-category rather than remediation order, so the work proceeds out of sequence and the early fixes are unobservable. Or the client works the easy items, because a list of forty presents no argument for why item thirty-one should come first.
An audit that has done its work returns one binding constraint, a short ordered sequence behind it, and the evidence for why that order and not another. Everything else it found belongs in an appendix.
The five evidence sets
A semantic SEO audit collects five evidence sets, and each one answers a different question about the property.
Coverage. Which tuples in the map the property renders, weighted by structural class, and where the gaps concentrate. Answers whether the property holds its subject.
Extraction. A sample of at least fifty headings, scored on whether the first sentence beneath each is an extractable triple carrying an exact value. Answers whether the coverage is usable.
Network. Orphans, click depth, bridge presence, the ratio of outer-to-core against core-to-outer links, and n-gram consistency. Answers whether the graph transports authority or leaks it.
Information gain. The proportion of core pages carrying at least one element the index does not already hold. Answers whether the property adds anything.
History. Months of continuous publication and the growth in ranking queries. Answers whether coverage has had time to be evaluated.
The five sets are the inputs to the Holistic Authority Score, and running them produces the score as a by-product rather than as a separate exercise.
→ The Holistic Authority Score: a computable measure of topical coverage
Separating symptoms from causes
Most findings in an audit are symptoms of a smaller number of causes, and a checklist cannot tell the two apart because a checklist records observations rather than relationships.
| Observed symptom | Usual cause | Not the cause, despite appearances |
| Two of your own pages trade positions for one query | Two URLs carrying the same contextual vector | Internal linking |
| Pages are indexed and rank for nothing | Coverage gap across the cluster, or absent information gain | Page-level optimization |
| Pages rank but are not cited in generated answers | Passages are not self-contained, or off-site standing is weak | Schema markup |
| Entry queries bring traffic that never converts | Missing middle-step pages, or bridges that are link strips | Calls to action |
| Coverage is good and crawl frequency is poor | Technical layer, or click depth beyond three | Publication volume |
| High-volume pages are live and competitive terms are lost | Volume-first publication order | Content quality |
| Coverage is broad and nothing ranks | No completed cluster, from parallel cluster opening | Insufficient page count |
| The same fact appears with two values across the property | No canonical value register | Editorial proofreading |
The right-hand column is the useful one. Each entry is the remediation a competent team reaches for first, and each is work performed on a symptom while its cause continues to produce more symptoms.
The diagnostic test is direct: if fixing the finding on one page would leave the same finding recurring on other pages, the finding is a symptom.
Resolving findings that conflict
Findings routinely imply contradictory priorities, and five precedence rules resolve them.
Rule one: technical precedes everything. Content that does not survive to the parser makes every other repair unobservable. A finding at the technical layer outranks a finding of any severity at any later stage.
Rule two: a cause is never remediated after its symptom. If the audit has identified a cause, the symptoms it produces are removed from the priority list entirely rather than ranked below it. Symptoms scheduled as separate work items are work performed twice.
Rule three: coverage precedes extraction. Immaculate sentences on a half-covered subject produce a property that is easy to read and not authoritative. Rewriting for extraction before closing coverage gaps optimizes documents whose main limitation is the absent documents around them.
Rule four: at equal precedence, prominence outranks volume. When two findings sit at the same stage, the one attached to the more prominent attribute is remediated first, regardless of the search demand attached to either.
Rule five: a finding recurring across three or more clusters is systemic. Systemic findings are remediated by changing the process that produces them, usually the content brief, rather than by fixing the instances. Fixing instances of a systemic finding guarantees the finding returns with the next publication run.
Rules two and five together typically collapse a forty-item findings list to between three and six real items. That collapse is the audit.
The verdict: one binding constraint
An audit’s verdict names one binding constraint, which is the single condition currently preventing the property from performing better than it does.
A binding constraint is not the largest finding or the most severe one. It is the finding whose removal makes the next finding matter. A property with a technical layer failure has that failure as its binding constraint even when its coverage gap is larger in points, because the coverage gap cannot be observed to close while the technical failure holds.
Naming one constraint is uncomfortable to deliver and it is the entire value of the exercise. A client given one constraint knows what to do on Monday. A client given forty findings knows only that their property has problems, which they knew before commissioning the audit.
The remainder of the sequence follows from the constraint rather than from severity, and the sequence is short: the constraint, then whatever the constraint was masking, then the next stage in order. Anything beyond six items is speculation about a property that will have changed by the time the sixth item is reached.
→ How to lower cost of retrieval: the six stages and the five layers
What a semantic SEO audit cannot tell you
A semantic SEO audit cannot tell you whether demand exists. The audit judges coverage against a map, and a perfectly covered map of a subject nobody searches returns a high score and no traffic.
A semantic SEO audit cannot tell you your off-site standing. Brand mentions, citations, and links influence both classical ranking and citation by generative systems, and none of them appear in any of the five evidence sets.
A semantic SEO audit cannot tell you whether the subject was worth choosing. Central entity selection is a judgment about the business, made before the map exists, and an audit inherits that judgment rather than testing it. A property centered on an entity it cannot monetize will audit well and earn nothing.
A semantic SEO audit cannot predict a timeline. It identifies the binding constraint and the sequence behind it. How long removal takes to show results depends on competition, domain history, and off-site profile, none of which the audit measures.
BRIDGE
The reasoning above is publishable because the reasoning is not the scarce part. The symptom map, the precedence rules, and the discipline of naming one constraint can all be read in ten minutes and applied to your own property this week.
What resists transfer is applying them to a property you did not build. An internal audit runs into the fact that the person auditing chose the contextual vectors, wrote the briefs, and set the publication order, which makes rules two and five almost impossible to apply honestly, because a systemic finding is a finding about your own process. A Holistic Radar authority audit builds the map where none exists, runs the five evidence sets, returns the Holistic Authority Score with its component breakdown, and names one binding constraint with the sequence behind it, so that building topical authority to rank starts from a decision rather than from a list.
→ Authority audits: one binding constraint, with the evidence behind it
SUPPLEMENTARY CONTENT
How often to audit
Audit quarterly, or after any publication run adding twenty or more pages. Auditing more frequently produces movement inside the noise of the heading sample, and the network layer degrades through accretion rather than through discrete events, which means it fails quietly between audits rather than visibly during them.
Whether a property without a topical map can be audited
A property without a topical map can be audited only after a map is built for it, because coverage requires a denominator. The map can be built retrospectively from the property’s existing subject rather than from a fresh strategic decision, which is faster than a full production build and sufficient for measurement.
An audit delivered against no map is a technical audit and a content audit presented together. That is a legitimate deliverable and it is a different one.
Whether to audit before or after a redesign
Audit before. A redesign changes the technical and structure layers and leaves coverage, information gain, and history untouched, which means a post-redesign audit reports the same substantive findings against a different template.
Auditing first also establishes whether the redesign addresses the binding constraint. Most redesigns do not, because most binding constraints are coverage or information gain and neither is a design problem.
What a good audit deliverable contains
A good audit deliverable contains the topical map with coverage marked, the Holistic Authority Score with its five components separated, the symptom-to-cause analysis showing which findings were collapsed into which causes, the binding constraint, the sequence behind it, and the full findings in an appendix.
The appendix matters. Removing findings from the priority list is a judgment, and the client should be able to see what was removed and why rather than trusting that nothing was missed.
→ Topical authority vs domain authority: what each metric measures