Before you start: what a useful checklist must meet
A self-respecting SEO audit is not measured by the number of points it reviews, but by the ones it discards. I have seen 250-point checklists that nobody finishes and that end up being used as an excuse to change nothing. The list that follows is designed to be executed in two real work sessions: one for on-page and one for off-page, with pass/fail criteria you can defend in front of a client or a product manager.
Three rules of use before getting into the substance:
- Every point must have a numerical threshold or a binary criterion. “Review titles” is not a point, it is an intention. “No duplicate title among indexable URLs in the same cluster” is.
- Separate findings from decisions. A report that mixes data and recommendations cannot be prioritized. Note the finding, then decide whether it goes into the sprint.
- Record what you have not been able to verify. An honest checklist includes a section for “out of scope” or “requires access to X.” If you do not have server logs, say so; do not infer crawl behavior from a third-party tool.
On-page block: checks performed with the page open
1. Match between intent and page format
Before looking at a single tag, compare the first three URLs in the top 10 with yours. If the SERP returns comparisons and your page is a product page, the problem is not on-page: it is architectural. This point is frequently failed, and fixing it usually involves creating a new URL, not editing the existing one.
Criterion: the dominant format in the SERP (guide, listing, comparison, product page, forum) matches that of your page, or there is a documented reason not to follow it (strong brand, mixed intent verified with CTR data).
2. Titles and metas: duplication and truncation
Export all titles from indexable URLs and group them by similarity, not by exact match. The usual mistake is having unique but templated titles: “Service X in City A,” “Service X in City B”… Google may rewrite them and you will lose control over CTR. According to Google’s official documentation on search snippets, the system can generate alternative titles when the original does not describe the page well, so uniqueness is not enough: the title must describe the specific content.
Criterion: no identical title among indexable URLs; length reviewed in pixels (not characters) for desktop and mobile; each title contains the page’s real differentiating term.
3. Hn structure consistent with content hierarchy
Check that there is a single H1 and that the H2s are not decorative. A pattern that appears often on corporate websites: H2s with design section names (“Our services”) and the H3 with the real topic. That wastes the semantic signal. Also check that no H2 is inside a component loaded by JavaScript that does not render in the initial HTML.
4. Minimum viable content per page type
There is no magic number of words. What can be audited is subtopic coverage: list the questions the SERP answers (people also ask, forums, autocomplete) and check how many your page covers. If you cover 3 out of 12, the problem is content, not keyword density.
Important nuance: on transactional pages (checkout, filtered categories) adding text usually worsens conversion. In those cases the criterion is structured data coverage and internal linking, not extra paragraphs.
5. Internal linking: the point almost nobody audits well
Do not count links: check whether the pages that generate revenue receive links from the pages with the most internal authority. Do this with a query of inbound links per URL and compare it with the list of URLs that convert (Analytics or the funnel itself). The typical mismatch: the blog links between articles and never toward product pages.
Criterion: each commercial URL receives at least one contextual link from a URL with demonstrated organic traffic.
6. Canonicals, pagination, and parameters
Check that the canonical points to the URL you actually want indexed and that there are no canonical chains. A frequent failure: canonical pointing to the version with a trailing slash while the sitemap lists the version without a slash. According to the official documentation on canonicalization, signals are consolidated, but contradictions between sitemap, canonical, and internal links reduce the reliability of the signal.
7. Structured data present and valid
Validate with the rich results tool and with the schema validator. Distinguish between “valid” and “eligible”: correct markup does not guarantee a rich result. Check that the declared type matches the visible content (do not mark Product on a category page).
8. Images and multimedia
Descriptive alt attributes on informative images, empty on decorative ones. Check the weight: a 2 MB image on the first screen ruins any on-page effort. Also check that relevant text is not embedded only in images.
Off-page block: what you cannot see from the CMS
9. Link profile: distribution, not volume
Audit the proportion of links to the home page versus links to internal pages. A profile with 80% of links pointing to the root domain suggests a brand strategy without topical depth. Also review anchor text: if all links use the same exact keyword, the signal is artificial.
Honest limitation: no third-party tool sees the complete profile. Link indexes are samples. Do not make disavow decisions based on a single source.
10. Toxic or purchased links: when to act and when not to
Most “toxic” links according to automated tools are not. Before disavowing, check manually: does the domain have real content? Is the link editorial or from a massive directory? Are there signs of a blog network (same owner, identical templates)?
A specific case that recurs: a services website inherits links from an old domain redirected 301 with a different topic. The tool flags dozens of toxic links that are actually residual and have not caused a problem. Disavowing without criteria can remove legitimate signals. Google’s documentation on invalid links is explicit: the disavow tool is a measure of last resort.
11. Brand mentions without links
Review mentions in the press, industry directories, and forums where you are named without linking. Converting them into links is public relations work, not technical SEO, but it goes into the checklist because many audits ignore it.
12. NAP consistency and local presence (if applicable)
If the business has a physical location, compare name, address, and phone number on the website, the business profile, and the main directories. Discrepancies are the most frequent local failure and are not detected from Search Console.
13. External channels that generate brand traffic
Off-page is not just links. Check whether there are mentions in video, podcasts, or communities where your brand is searched afterward. A spike in brand searches without a campaign can be explained by an external mention; if you do not detect it, you will attribute the growth to the last on-page change you made.
How to prioritize findings without dying in the attempt
Classify each failed point into three boxes:
- It breaks indexing or crawling: fix it this week. Includes persistent 5xx errors, accidental noindex, robots.txt blocking resources, cross-canonical between different pages.
- It affects relevance or CTR: goes into the next content cycle. Titles, subtopic coverage, structured data.
- It is strategic and slow: internal linking, brand mentions, link profile. Requires a calendar, not a sprint.
A common methodological error: prioritizing by a tool’s “estimated impact.” Those estimates are based on search volume and average position, not on your business model. Always cross-reference with revenue or leads per URL.
What this list does not cover
This check does not replace a deep technical audit (logs, rendering, crawl budget) or an intent analysis at scale. Nor does it solve content problems that require product knowledge. If the site has more than 50,000 URLs, manual sampling of titles and Hn stops being reliable and you need automation with validation by segments.
And a final warning: no checklist guarantees rankings. It serves to remove friction and leave the site in a position to compete; the result also depends on the competition, accumulated authority, and factors you do not control.