Perform an SEO Audit: The Order of Operations, from Collection to Action Plan

What you settle before starting

The decision to audit has been taken: what remains is to carry out the work without spending three weeks on it or producing four hundred alerts that nobody will act on. Two things are settled before opening a single export — what the audit has to demonstrate, and what it covers. Skipping them amounts to sorting anomalies as they come. For the upstream framing — what an audit covers, when it is triggered, how often it comes back — see the SEO audit.

The objective the audit has to demonstrate

You do not run an audit to produce a checklist: you run it to answer an explicit, shared and measurable objective. Aiming at “position” is no longer enough anyway, since the proportion of searches without a click (zero-click) reaches 60% (Semrush, 2025). A usable objective looks more like this:

• increase organic clicks on a set of landing pages;
• improve CTR on queries with high impressions;
• strengthen conversions from organic, prioritizing the pages that already convert;
• secure the indexing of “business” pages — offers, categories, pillar content.

These objectives sort themselves into a hierarchy as soon as you look at where the clicks are. The click-through rate on the first organic position (desktop) is 34% (SEO.com, 2026), while the click-through rate on page 2 of the SERPs is 0.78% (Ahrefs, 2025). Moving a page up from page 2 to the bottom of page 1 is therefore not worth the same as moving it from fifth to second place: it is this kind of gap that decides between two projects. The SEO statistics gather these benchmarks with their source and their year.

The scope and the access

To avoid the “titanic audit”, define what you are auditing: the scope — main domain, subdomains, directories, pages to exclude such as the admin area or internal search — and the depth, that is, a broad view completed by a focus on the key pages. The most pragmatic approach is to select a shortlist of 20 to 50 high-stakes URLs first — organic entries, best conversions, sharpest drops, and strategic pages that are still invisible — then widen it if necessary. That last group is the one most often forgotten: limiting the audit to pages that are already visible or converting hides precisely the most costly blockages, since a page that is blocked, deindexed or never crawled appears in no performance export.

Without reliable data, the audit becomes an opinion. Two sources form the pre-diagnosis foundation, and the method uses them at every step. Search Console gives performance (clicks, impressions, CTR, position), indexing status and manual actions. Google Analytics gives organic landing pages, engagement and conversions. Prepare the exports before you start, over 28 days compared with the previous period: re-running an extraction mid-analysis loses comparability. This foundation is not enough to audit with, because it says what is going wrong and rarely why: you have to add at least a crawl of the site — the only way to obtain the inventory of URLs, the statuses, the directives and the real internal linking, including pages that are not visible — and an inspection of the rendering, which checks what an engine sees once the JavaScript has run. Inbound link data is added depending on the scope of the assignment, when popularity is among the questions asked.

Step 1 — start from real performance, not from the list of anomalies

Start with real performance, not with a list of anomalies. This is the rule that separates a usable audit from a tool export: anomalies are counted in the hundreds, the pages that carry the value are counted in dozens, and it is the latter that decide what you look at.

Spot the pages that matter

In Search Console, export the pages and the queries that combine:

• high impressions, a sign of coverage and potential;
• significant clicks, which prove interest and intent;
• positions close to the top 10, where the gain is fastest;
• and, on the Analytics side, those that contribute to conversions, even at lower traffic volume.

Crossing these four filters with the framing shortlist gives the working list for the whole audit, with one reservation: these filters only see what already appears in the reports, and strategic pages that are still invisible are added to it by hand, from the crawl. A page that is not on it will not be examined this time round: that is not an oversight, it is the decision that protects the time available. This list will be used again at step 4, when you have to justify why one action comes before another.

The quick gaps and the testable hypothesis

The most profitable opportunities often sit on queries that are already visible. Look in particular for pages in positions 4 to 15 with high impressions: they most often deserve work on structure, proof and internal linking. And pages with a low CTR at a stable position: the problem then frequently comes from the snippet, or from a results page that has become richer. On this second case, an optimized meta description can improve CTR by 43% (MyLittleBigWeb, 2026) — to be read as an order of magnitude, not as a guarantee page by page.

For each key page, then write a testable hypothesis, along the lines of: “impressions are falling because the page is excluded, canonicalized or redirected”; “the position is stalling because the content does not cover the intent”; “the CTR is low because the title is generic and does not promise a clear answer”. The point is to connect a symptom to a likely cause, then to a measurable action. A written hypothesis gets checked at the next step; an unwritten hunch turns into a project whose purpose nobody can state any more.

Step 2 — the checks that block visibility

The hypotheses from step 1 are now confronted with what the site actually does. The principle: only check on the working list, and stop at the finding. You record, you qualify, you fix nothing yet.

Indexing and coverage

In Search Console, examine the reasons for non-indexing and compare them with your shortlist of high-stakes pages. Not every URL is meant to be indexed — admin pages, redirected pages — but an excluded “business” page has to trigger an action. Use URL inspection to confirm the status, then request indexing after the fix and validate the problem once it has been handled.

This check names the blockage, it does not explain it. When the finding calls for a rule — crawl directives, statuses, canonicalization, JavaScript rendering, orphan pages — it is the technical SEO audit that gives the procedure to follow.

Content and structure of the priority pages

On the pages in the working list, check what influences understanding and the click: an explicit, non-duplicated, intent-driven title; a readable structure in H2/H3, with a clear answer from the start; proof elements — examples, attributed figures, definitions. Three faults come up often: a title that describes the brand rather than the answer, a page that makes its answer wait three screens, and a title promise the body does not keep. Each is noted as a finding with its URL, without writing the fix at this stage.

Internal linking, duplication and cannibalization

Internal linking serves both crawling and hierarchy. Look for high-stakes pages that are too deep, well-performing pages that do not redistribute towards your strategic pages, and anchors that are too vague or over-optimized. One practical rule often applies: add a few contextual links, in the body text, to pages in positions 6 to 20, without forcing the anchors.

Two signals are recorded along the way, without being settled. Technical duplication: several versions of the same URL accessible, which dilutes the signals. Editorial cannibalization: in Search Console, the query “changes” URL, or positions alternate between two pages. These are audit findings; designating the page that keeps the topic and deciding what to do with the other belongs to the semantic audit, which carries the rules for choosing between creating, optimizing and consolidating.

Step 3 — interpret without picking the wrong fight

This is the step where an audit most often loses its way: the findings are there, and the temptation is to treat everything at the same level. Interpreting means deciding what a piece of data proves, then what deserves to enter the action plan.

What Search Console confirms, and what it does not prove

Search Console is excellent for connecting pages, queries, impressions and clicks. Use it to confirm that a page is indeed visible, that it gets clicks, and that an indexing problem really affects strategic pages. On the other hand, on its own it does not “prove” the single cause of a drop: it shows a gap, not its origin.

The consequence is simple. A hypothesis from step 1 is confirmed when the data and the field finding say the same thing — the page really is excluded, the title really is generic. When they diverge, the hypothesis is not wrong for all that: it stays open and does not become an action. An action built on an unestablished cause produces the fixes that nobody will be able to judge, six months later, as having served any purpose.

The false-positive filter and the cross-reading

An audit can surface hundreds of “noisy” points. The operational filter is simple: an anomaly is a priority if it affects (1) organic landing pages, (2) pages that convert, (3) a strategic page that is still invisible, or (4) a whole template, through sheer scale. An anomaly that ticks none of the four boxes does not enter this round's action plan. That is the rule to invoke when the production team asks why one ticket comes before another.

Three cross-readings are then enough to qualify what remains:

• stable impressions and falling clicks: a likely CTR problem — the snippet, or a richer results page;
• falling impressions: a likely problem of indexing, canonicalization or loss of relevance;
• stable organic traffic but falling conversions: content or journey that is less persuasive, or a poor match between the query and the offer.

Step 4 — from finding to prioritized action plan

The expected output of all of the above comes down to one sentence: a few strong decisions, correctly ordered, rather than a massive unsorted backlog. Three moves lead there — score, order, write.

Score: SEO impact, business impact, effort, risk

To prioritize without endless debate, give each action a simple mark, from 1 to 5, on four axes:

• SEO impact: crawling, indexing, ranking, CTR;
• business impact: leads, revenue, strategic pages;
• effort: time, dependencies, release cycle;
• risk: regression, side effect, traffic loss.

The scoring does not have to be accurate to the decimal: it has to be done in front of the people who will execute. A developer who scores the effort of an action himself will not contest it when the ticket comes, and an acquisition manager who scores the business impact owns the order that comes out of it.

Order: quick wins, structural projects, continuous optimizations

The marks translate into three batches, which do not run at the same pace:

• Quick wins: pages already close to the top 10, abnormally low CTR, internal linking to strengthen on a few hubs.
• Structural projects: content consolidation, rebuilding a template, global URL rules.
• Continuous optimizations: gradual enrichment, iterations on the editorial plan, tracking SERP changes.

The classic trap is to launch all three batches in the same week. Quick wins produce their effect fast and can be measured in isolation; structural projects affect whole templates and blur the measurement of everything else if they start at the same time. Sequencing preserves the readability of what you will have produced.

Write an executable action

A recommendation that does not become an action will not be executed. Each one states the scope (URL, template, directory), the owner (SEO, content, development) and a validation criterion, that is, the proof expected — an improvement in indexing in Search Console, or a rise in clicks on a comparable group of queries. Three examples written end to end:

• “On [URL], reword the title to reflect the main intent observed in Search Console, then measure CTR over 28 days against the previous period.”
• “Attach [URL] to the [category] hub via 3 contextual links from high-traffic pages, with varied anchors, then check the rise in impressions and clicks.”
• “Rework [URL] to cover the missing sub-topics recorded at step 2, then check the average position on the target queries.”

The short audit: what you check when you have only an hour or two

The full sequence takes several days. When time is short, it comes down to a routine of ten checks, which fits into one to two hours on a mid-sized site and covers the four steps in compressed form:

• 1. Identify the top 10 to 20 organic landing pages and their trends.
• 2. Compare clicks and impressions over 28 days against the previous period.
• 3. Spot 10 non-brand queries in positions 6 to 15 with high impressions.
• 4. Check the important pages excluded from the index.
• 5. Check the errors affecting high-stakes pages, or their internal links.
• 6. Spot the pages with a low CTR despite a decent position.
• 7. Check the consistency of the URL versions and the expected redirects.
• 8. Check 3 to 5 “money” pages: title, Hn hierarchy, direct answer, outbound internal linking.
• 9. Identify a likely cannibalization — a query that switches from one URL to another.
• 10. Produce a mini-backlog of 10 actions at most, prioritized.

Beyond that, only two levels, and they do not change the sequence: a standard audit applies it to a shortlist of pages and takes a few days; an in-depth audit applies it template by template and takes several weeks. When you have only an hour, checks 7 to 9 are the first to go: the most manual checking for the fewest decisions. Search Console and Analytics give the pre-diagnosis in both cases; the crawl and the inspection of the rendering remain necessary whatever the level, and the choice of SEO audit tools arises as soon as volume exceeds what can be checked by hand.

Four mistakes distort the diagnosis, whatever duration is chosen:

• Auditing without an objective: you pile up findings but no decision follows.
• Treating everything at the same level: a micro-anomaly on a non-strategic page must not block a structural fix.
• Confusing correlation with causation: a drop can come from a richer SERP, not from a problem “on the page”.
• Not planning the validation: without acceptance criteria, you declare “fixed” what is not.

How much time to spend, and what you deliver at the end

There is no single duration. The principle: the larger the site and the more executable the expected write-up — a backlog with its validation criteria — the more time it takes. Break it into four blocks: collection and exports; analysis focused on high-stakes pages; prioritization; write-up. The first grows with the number of sources and the state of the access, the second with the number of distinct templates, the third hardly at all. It is the fourth that overruns most often, because it includes the internal coordination time between content, development and marketing, often underestimated: plan for it at framing rather than at the end of the road.

These same factors make the budget vary when the audit is outsourced: scope covered, depth of analysis, level of write-up. Depending on the case, billing takes the form of a fixed fee on a scope defined in advance, time and materials, or a mix — a fixed fee for the diagnosis, time and materials for supporting the fixes. Knowing how you will be billed, and on exactly which scope, is settled before starting. Price is not a reliable indicator of quality: what is reliable is how applicable the recommendations are and how clear the action plan is.

At the end, three deliverables at a minimum: a summary (major problems, opportunities, risks), an export of the data analysed, so that the findings can be reread, and a prioritized backlog with owners, deadlines and expected proof. What separates a usable backlog from a list of recommendations comes down to three refusals. No recommendation without proof: a finding with no URL and no data cannot be checked. No action without an owner: it will stay in the document. No action without a validation criterion: without one, you will declare “fixed” what is not, and the next audit will start from scratch. These three refusals avoid the two most frequent pitfalls, the “black box” audit and the backlog that cannot be implemented. Replaying the same sequence after every release, without redoing the collection by hand, is precisely the task covered by the audit and mapping module.

FAQ: frequently asked questions about performing an SEO audit

What is an SEO audit, in practical terms?

It is a structured analysis of what influences your visibility — Search Console and Analytics data, indexing consistency, content, internal linking, popularity — whose expected output is a prioritized and verifiable action plan. A document that lists anomalies with no hierarchy and no proof is not one.

Which practical steps should you follow so as not to forget anything?

A four-part sequence: (1) start from real performance to designate the pages to work on, (2) check on those pages only what is blocking visibility, (3) interpret while filtering out false positives, (4) turn the findings into a prioritized backlog. Validation after implementation closes the cycle.

What can you check in 30 to 60 minutes (quick analysis)?

The core of the ten-check routine: landing pages and trends, comparison of clicks and impressions over 28 days, non-brand queries in positions 6 to 15, important pages excluded from the index, errors on high-stakes pages, low CTR at a decent position. In an hour, you drop the most manual checks and come out with a mini-backlog of ten actions.

Which elements should be checked first in an audit?

What can block or amplify the visibility of high-stakes pages: indexing, URL consistency, major errors, duplication and cannibalization, internal linking towards strategic pages, then CTR and content structure. In that order: an unindexed page gains nothing from a better title.

Which free tools can you use with no software budget?

Search Console and Google Analytics form the pre-diagnosis foundation, and everyone already has them. They replace neither a crawl nor an inspection of the rendering: crawlers generally offer a free version capped on the number of URLs, enough on a small site, and Search Console's URL inspection shows the rendering page by page. Inbound link data is added depending on the scope of the assignment.

Which deliverables should you expect at the end of an SEO audit?

A summary, the data analysed as exports, and above all a list of prioritized actions — with impact, effort, risk, owner and deadline — together with their validation criteria. A recommendation with no proof, no owner or no validation criterion should not be accepted as a deliverable.

How do you interpret the results without overreacting?

Cross the signals: an alert is only a priority if it affects organic landing pages, pages that convert, a strategic page that is still invisible or a whole template. Remember that Search Console shows a gap, not its cause: as long as the data and the field finding diverge, the hypothesis stays open and does not become an action.

How do you prioritize the actions after the audit?

Score each action from 1 to 5 on four axes — SEO impact, business impact, effort, risk — then group them into quick wins, structural projects and continuous optimizations. Avoid aiming at “zero alerts”: aim at a few strong decisions, correctly ordered.

How long should it take, depending on my context?

It depends on the volume, the number of distinct templates and the level of write-up expected. Count one to two hours for the short routine, a few days for a standard audit on a shortlist of pages, several weeks for an in-depth audit template by template. Always plan for the internal coordination time.

How much does an SEO audit cost in 2026?

The budget varies with the scope covered, the depth of analysis and the level of write-up expected. Billing takes the form of a fixed fee on a defined scope, time and materials, or a mix of a fixed-fee diagnosis and support on a time and materials basis. Price is not a reliable indicator of quality: the right benchmark is how applicable the action plan is.

What do you do if everything “looks fine” but positions are not moving?

Go back to the pages and the queries: check the match between the intent and the page, the depth and credibility of the content, the internal linking, then the CTR. If impressions and positions are stalling together, look for a coverage gap — sub-topics expected and missing — or cannibalization between two of your pages.

How do you validate that the fixes had a real impact?

Define a comparable before/after (same periods, same segments), then track in Search Console the impressions, clicks, CTR and positions on the target queries, and in Analytics the quality of the visits and the conversions of the pages concerned. Document the deployment dates: without them, attribution is lost.

Continue reading

• The audit is done and you still have to write it for three different audiences: an SEO audit example gives the report's chapters, the recommendation table and its columns.
• You do not have the time to run the sequence yourself and have to frame what you are commissioning: what to require from a supplier on an agency SEO audit, deliverables and acceptance included.
• The actions are delivered and the one-off validation criterion is no longer enough: SEO tracking details KPIs, annotations and the monitoring of gains over time.
• Your visibility plays out in one area rather than across a whole country: the local SEO audit changes what gets checked — local signals, NAP consistency, reviews, business profile.