An Audit Serves to Decide, Not to Produce a Document
Most SEO audits delivered on the market are formatted tool exports. Sixty pages of screenshots, a list of three hundred anomalies with no hierarchy, and a client who closes the document without knowing where to start.
A useful audit produces something else: a short list of actions ranked by effort-to-impact ratio, with what each costs, what it delivers, and who executes it. It runs ten pages and reads in twenty minutes.
This difference comes down to method. A tool detects anomalies—it does not prioritize them and does not know your business. A missing title tag on a legal notice page and a two-second server response time appear in the same table, even though one has no importance and the other caps the entire site.
The order of verification is not indifferent. Certain blockages cap everything else: a non-crawlable page will not rank, regardless of content quality. The twelve points below are ranked in the order we address them, from blocking to perfectible.
Point 1: Actual Site Indexation
The first question is not how your pages rank, but which ones Google knows. The answer often surprises.
The Search Console coverage report shows how many pages are indexed, how many are excluded, and why. A significant gap between the number of pages on the site and the number indexed signals a structural problem that makes any other optimization useless.
The most common exclusion reasons are a canonical tag pointing elsewhere, a blocking directive inherited from a test environment, a page considered too thin, or an address simply never discovered for lack of an inbound link.
The check takes ten minutes and it conditions everything else. Do it before opening a keyword analysis tool.
Point 2: The Exclusion File and the Sitemap
Two tiny files capable of making an entire site disappear. The first tells robots what they must not crawl, the second lists what they should know.
Classic errors on the first are a total block forgotten after launch, blocking of resources needed to render pages, or exclusion of an entire directory out of excessive caution. These errors are rare but catastrophic, and they take thirty seconds to verify.
On the second, defects are more ordinary and more frequent: a sitemap listing redirected or error addresses, forgetting entire sections, or not being declared to Search Console. A clean sitemap accelerates discovery of new content; a dirty one sends contradictory signals.
Point 3: Redirects and Chains
On an old site, redirects pile up through successive redesigns. One address redirects to a second, which redirects to a third, which ends on an error page.
Each additional hop dilutes transmission and slows loading for the visitor. Beyond three hops, some robots give up. The fix is to point each old address directly to its final destination, eliminating intermediaries.
The second check concerns internal links pointing to redirected addresses. They work, so no one corrects them, but they impose an unnecessary hop on every visit and every crawl. Updating them to the final address is mechanical work with good returns.
Point 4: Server Response Time
The time the server takes to return the first byte conditions everything that follows. The benchmark communicated by Google sits below eight hundred milliseconds, and many sites are far from it.
Three causes dominate: saturated shared hosting, an old server language version, and absence of caching. These three points are addressed infrastructure-side and produce immediate effect across all pages, unlike front-end optimizations that must be applied template by template.
It is also the factor that determines crawl budget. A slow server reduces the number of pages the robot covers per session, which slows indexation of new content on large sites.
Point 5: Field-Measured Loading Experience
The three experience indicators are measured on real visits, not on a simulation. A site can display a flattering lab score and fail on field data, because the latter integrate modest devices and degraded networks.
The check is done in Search Console, which groups addresses by template and flags failing groups. This group reading is essential: when two hundred pages appear in the same group, the fix is made once in the template.
The detail of the three metrics, their thresholds, and applicable fixes is covered in our article on Core Web Vitals. For the audit, what matters is knowing which templates fail and on which metric, to direct the work.
Point 6: Architecture and Depth
An audit must produce a site map as the robot sees it, with each page’s click depth and number of internal inbound links.
Two anomalies emerge systematically. Strategic pages buried four or five clicks deep, receiving almost nothing. And secondary pages very well served, often because they appear in the menu or footer.
The fix is a matter of link mechanics rather than restructuring, and it is detailed in our article on internal linking. When the problem comes from the structure itself, you must go back to architecture, a subject covered in our semantic silo method.
Point 7: Title Tags and Descriptions
The check is basic and it almost always reveals work. Three anomalies to look for: missing titles, titles duplicated across multiple pages, and titles truncated for lack of controlled length.
Duplication is the most significant. Two pages carrying the same title signal to the engine that they address the same subject, which triggers arbitration from which neither emerges winning. On automatically generated catalogs, this defect sometimes affects hundreds of pages.
Descriptions do not influence ranking but they influence clicks. An absent description lets the engine generate an excerpt that is not always flattering. On pages receiving traffic, rewriting them is quick action with immediate returns.
Point 8: Heading Structure Within Pages
One main heading per page, subheadings hierarchized without level skipping, and labels containing subject vocabulary rather than hollow formulas.
The observed defects are always the same: multiple main headings on the same page, often because the logo and article title both carry this level. Subheadings used for their appearance rather than their function. And generic labels like “Our Services” that carry no query.
This check is quick and its effect is measurable, because section labels directly participate in the engine’s understanding of the subject.
Point 9: Duplicate Content and Canonicals
Internal duplication is more frequent than external duplication and easier to fix. It comes from versions with and without trailing slash, URL parameters added by filters, printable versions, and poorly managed pagination pages.
The canonical tag resolves most of these cases by designating the reference version. It still must be correct: a canonical pointing to the homepage on all site pages is a classic error that makes all interior content disappear.
The check consists of verifying, on a sample from each template, that the canonical points to the expected address. The complete subject, with cannibalization cases between distinct pages, deserves separate treatment.
Point 10: Images and Media
An often-neglected area even though it represents the majority of page weight. Three checks suffice.
Format first: serving WebP or AVIF rather than JPEG reduces weight by thirty to fifty percent at identical quality. Dimensions next: a three-thousand-pixel image reduced by the browser in an eight-hundred-pixel container wastes bandwidth on every visit. Attributes finally: width and height declared to avoid shifts, alternative text filled in for accessibility and image search.
On sites we audit, this single area regularly represents half the available performance gain, for purely mechanical work.
Point 11: Structured Data
The check addresses three questions. Does markup exist, is it valid, and is there only one source generating it.
This last question catches the most frequent error: an extension and a template each produce their declaration, the two contradict each other, and the result becomes unpredictable. Verification is done on the live address and not on the written code, because the gap between the two is precisely where the problem hides.
The complete method and useful types by sector are detailed in our JSON-LD markup guide.
Point 12: Credibility Signals
Last point on the list and first differentiation factor in contested markets. It is measured by no tool and verified by hand.
Are contents signed by an identifiable person with a dedicated page. Are legal pages complete and current. Does a verifiable physical address appear on the site. Do contents carry a publication and update date. Is what the site claims consistent with what external sources say about it.
These checks fall under the framework described in our article on E-E-A-T, and they weigh particularly on sectors where the decision involves money or health.
The Necessary Tools, and Those That Are Not
A complete audit is conducted with three free tools and one paid, and that suffices for the vast majority of sites.
Search Console provides indexation, coverage, search performance, experience signals, and markup errors. It is the source of truth, because it exposes what Google actually holds on your site.
A performance measurement tool gives field assessment and lab diagnosis on a given address. The browser itself, with its development tools, allows inspection of network, rendering, and console errors.
The only truly paid tool needed is a complete site crawler, which covers all addresses by following links and produces the map. It is what reveals redirect chains, orphan pages, and real depth.
What is not necessary: all-in-one suites at several hundred euros monthly, useful for competitive monitoring but superfluous for technical diagnosis. An audit is not paid for in subscriptions, it is paid for in analysis time.
Prioritizing Results: The Part That Makes the Audit
Detecting is mechanical, prioritizing requires judgment. We rank each anomaly on two axes.
Impact first: how many pages are affected, and do these pages carry traffic or revenue. An anomaly on three hundred product sheets does not weigh like the same anomaly on a legal notice page.
Effort next: does the fix involve a setting, a template modification, or a rebuild. A one-hour fix affecting two hundred pages comes before a three-day fix affecting five.
The intersection of the two produces four families: do immediately, schedule, consider later, ignore. This last category is the one most often missing from market audits, yet it is the most useful: saying what does not deserve correction saves as much time as saying what must be corrected.
Read Search Data Before Concluding
A purely technical audit misses the richest information: what Google already knows about your site and what users actually search to reach it. This data is free and underexploited.
Queries Bringing Impressions Without Clicks
The performance report lists queries on which your pages appear. Those generating many impressions and very few clicks signal a gap: either your page ranks too low, or its title and description do not entice. The second case is fixed in an hour and produces immediate gain.
Pages Ranking on the Second Page
An average position between eleven and twenty means the page is recognized as relevant but insufficient. These are the most profitable targets of an action plan: the work required to cross the threshold is far less than to make a nonexistent page emerge.
The fix generally combines three interventions: enrich content on aspects that better-ranked results address, strengthen internal linking to this page, and verify that no other site page targets the same query.
Queries Captured by the Wrong Page
It frequently happens that a commercial query is carried by a blog article rather than by the corresponding service page. Traffic arrives, but on a page not meant to convert. The fix consists of adjusting linking and sometimes labels, so Google designates the right page.
Auditing the Competition, Briefly
An audit turned exclusively inward produces an incomplete diagnosis, because SEO is played in relative position. Three observations suffice, and they take an hour.
Identify who occupies the top three positions on your main commercial queries. Look at the structure of their pages: length, subheadings addressed, presence of proof and figures. Measure their loading time with the same tool as yours.
These three observations give a realistic target. If ranked pages run four thousand words with proprietary data and yours runs six hundred, the gap is not technical. If they load in one second and yours in four, it is.
This observation also avoids the opposite error, which consists of over-investing in an already-won market. A company already occupying first position on all its commercial queries does not need a technical audit, it needs to broaden its semantic field, which is different work.
The Audit Report: What It Must Contain
Form matters because it determines whether the document will be used or filed.
A summary page first, with the three main blockages and the expected effect of their correction. It is the only page the decision-maker will read; it must be self-sufficient.
The action plan next, in table form: action, affected page or template, estimated effort, expected impact, responsible party. Ranked from most profitable to least profitable, with an explicit separator line marking the threshold below which corrections are not worth the time they cost.
Technical detail finally, in appendix, for the person who will execute. This detail need not appear in the document body, where it drowns decision information.
An audit delivered without a restitution meeting loses half its value. Arbitrations are discussed, priorities depend on constraints the provider ignores, and one hour of exchange avoids six months of poorly executed plan.
What a Technical Audit Does Not Cover
Three subjects escape technical diagnosis and deserve announcement to avoid misunderstandings.
The relevance of commercial positioning. A technically flawless site targeting queries without commercial value will produce traffic and no clients. This question falls under keyword research, a distinct exercise.
Editorial quality. A tool detects a page that is too short; it does not detect a hollow page. Content evaluation requires human reading by someone who knows the subject.
External authority. Inbound link profile is measured with other tools and worked on long cycles. A technical audit mentions it; it does not address it.
These three components compose, with technique, a complete organic SEO service. The audit is the starting point, not the scope.
How Often to Audit
A complete audit is justified at three moments. Before committing an SEO budget, to know where the money goes. After a redesign, to verify the migration went well. And when an unexplained drop occurs.
Between these moments, light monitoring suffices: monthly consultation of indexation coverage and experience signals, quarterly verification of architecture and redirects. This monitoring is part of what a maintenance contract integrating the SEO component covers.
Auditing every month brings nothing. Corrections take several weeks to produce their effect, and measuring too often produces noise that is wrongly interpreted as signal.
Key Takeaways
An audit is judged on its conclusion, not on its thickness. If it does not end with a list of fewer than twenty actions ranked by effort-to-impact ratio, with an explicit category of what should not be corrected, it has not done its work.
The order of the twelve points is not decorative. Indexation, crawl files, redirects, and server response time cap everything else: as long as they are not resolved, working on titles or markup produces nothing measurable.
The first four points are verified in one hour with free tools, and they often suffice to identify the main blockage. Start there before committing anything, including before ordering a complete audit.
