Search for a checklist and you will be handed somebody’s inventory. Ninety items, a hundred and twenty, two hundred and six, arranged by category and numbered so the count is the first thing you see. Item count is the easiest thing for a checklist to compete on and the least useful thing for the person holding it, because a list nobody finishes is a list that got abandoned somewhere in the middle, on whatever happened to be item forty.
This is a full SEO website audit checklist built the other way around. Every item on it has been graded by one question, which is what actually happens if the check fails. Some failures take a site out of the index. Some decide which of your own pages wins a query. Some change nothing at all that anyone has been able to demonstrate, and those have been separated out and named rather than quietly dropped, because they will be on every other list you read and you deserve to know why they are not on this one.
What a checklist is for, and what it is not
A checklist is a reference artifact. You hold it while you work and you tick things. That is a different object from a procedure, which tells you what order to do the work in, where to capture a baseline, and how to tell a result from noise afterward. If what you need is the procedure, our walkthrough of how to measure SEO performance and fix what you find is the piece that covers the sequence, the baseline and the re-measure date. This page is the list you hold while you do it.
The distinction matters because it decides what belongs here. A procedure can afford to be long, since you only meet each step once and in order. A list gets read all at once, scanned, and judged on whether it looks finishable. Length is not a neutral property of a list. It is the thing that determines whether anyone gets to the end.
The two objects get conflated because most published checklists are procedures wearing numbered bullets. You can tell by whether the items assume an order. If item twelve only makes sense after item eleven, you are reading a walkthrough that has been reformatted to look scannable, and it will behave like a walkthrough when you try to use it, which is to say it will stall the moment you skip something.
The three questions every check has to answer
Before an item earns a place, it has to survive all three of these. Most published items fail at least one.
Does failing it change what a search engine can do with the page? Not whether it is good practice, not whether a tool flags it. Whether the page behaves differently in search because of it. A great many checklist items are style preferences that have been promoted to defects by a scoring tool that needed something to deduct points for.
Can you tell it has failed? Some things are genuinely important and genuinely unobservable from where you are sitting. If the only honest answer is that you would need data you do not have, the item belongs in a different section of the list, not alongside things you can settle in thirty seconds.
Can you do anything about it? A check that surfaces a problem you have no route to fixing is an anxiety generator. It still belongs on the list sometimes, because knowing is worth something, but it should be labeled rather than filed next to items you can close this afternoon.
Tier one, the checks that can make a site invisible
There are not many of these and that is the point. They are binary, they are fast, and every one of them can single-handedly undo everything else on the list. Run them first, and if any of them fails, stop and fix it before you look at anything else, because nothing below this tier can matter until these pass.
Google’s own starter guide makes the same move before it recommends anything. Its instruction is to check whether your content has been found at all first, and it adds a parenthesis that most checklists would never allow themselves, that maybe you do not need to do anything. That is the correct order of operations and it is the order this list uses.
Check that your robots file is not disallowing what you want indexed. Check that your important templates do not carry a noindex directive, which is the single most common way a staging setting reaches production and quietly removes a site from search. Check that canonical tags point at the page itself rather than at a different page, since a canonical aimed somewhere else is an instruction to count this page as that one. Check that your key templates return a success status rather than an error or a redirect. Check that the resources needed to render the page are not themselves blocked, because a page whose content only appears after a blocked script runs is, from the outside, an empty page.
Five checks. On most sites all five are clean and you are done in ten minutes. On the sites where one is not clean, you have just found the entire problem, and every one of the ninety items you would have worked through instead would have been noise on top of it.
Tier two, the checks that decide which page wins
Assume the site is reachable and indexable. The next question is not whether you rank but which of your pages is competing, because the most common failure at this tier is a site competing with itself and reading the result as weak authority.
Check that each query you care about has one page that is obviously the answer to it. Where two pages could both plausibly serve a query, one of them is going to win and it will not reliably be the one you chose. Check what your internal links say about which page that is, since the page you link to most from relevant context is the page you have nominated whether you meant to or not. Check that titles describe the specific page rather than the site, which is the difference between a title that can win a query and a title that is a brand statement.
There is a check here that almost no list carries and it is one of the most decisive available to you. Look at the queries a page actually receives against the query you built it for. Search Console will show you, per page, what people searched before they arrived. When those two lists disagree, you have learned something no audit tool can tell you, which is that the page is understood as an answer to a question you were not asking. That failure is observable, it is fixable, and it explains a great many pages that rank respectably and convert nothing, because they are meeting an audience that came for something adjacent.
The fix is rarely more content. It is usually deciding which of the two questions the page is for, then letting a different page take the other one. Adding paragraphs to serve both is how a clear page becomes an ambiguous one.
Duplication belongs here too, and it is worth being accurate about what it costs. Google’s documentation states that having duplicate content on your site is not a violation of its spam policies, while noting it can be a bad user experience and can waste crawling resources on URLs you do not care about. That is a real cost and a much smaller one than the penalty language that circulates about it. Treat duplication as a clarity problem, which is what it is, rather than as a threat.
Deciding which of two overlapping pages should survive is a content judgment rather than a technical one, and it has its own method. Our guide to what to keep and what to delete in a content audit covers the four verdicts and why three of them are not delete.
Tier three, run once and leave alone
These are real and they are not urgent. They reward being done correctly once and they punish being fiddled with quarterly.
Confirm a sitemap exists, is reachable, and lists pages you actually want indexed rather than every URL the system can generate. Confirm redirects go straight to their destination instead of through a chain. Confirm structured data, where you use it, describes what is genuinely on the page. If you serve more than one country or language, confirm your hreflang annotations point at each other in both directions, which is where almost all hreflang problems live.
Site structure sits in this tier and the temptation is to promote it. Google’s guidance is explicit about not doing that. Its starter guide advises against dropping everything to reorganize your site, noting that while the suggestions can help long term, particularly on a larger site, search engines will likely understand your pages as they are right now regardless of how the site is organized. A restructure is a large, risky project that a checklist should not be able to trigger on a whim.
The busywork tier, and why it survives
Here is the section that other checklists do not have, and the reason this one is shorter than theirs.
The keywords meta tag. Google’s documentation states plainly that Google Search does not use the keywords meta tag. It has been saying so for a very long time. It is still checked by tools, still reported as a missing element, and still added to sites by people who were told it was an oversight.
Keyword density. There is no published target because there is no mechanism that a target would describe. What the documentation does address is the failure mode at the extreme, warning that excessively repeating the same words, even in variations, is tiring for users and that keyword stuffing is against its spam policies. That is guidance about a ceiling and a reader, not a percentage to hit.
Keywords in the domain or the URL path. Worth stating precisely, because the belief is so durable. The guidance is that from a ranking perspective, the keywords in the name of the domain or URL path alone have hardly any effect beyond appearing in breadcrumbs. Sites are still bought for this.
Crawl budget, on almost every site that checks it. Google’s crawl budget guide opens by telling most readers to leave. Its opening instruction is that if your site does not have a large number of pages that change rapidly, or if your pages seem to be crawled the same day they are published, you do not need to read the guide. Crawl budget appears on general checklists constantly and applies to a small minority of the sites running them.
Third-party authority scores. This one is our judgment rather than Google’s, and you should weigh it that way, because unlike the five items above it rests on no documentation at all. The mechanism is the whole of the argument. A vendor score is that vendor’s own estimate, calculated from that vendor’s own index of the web, so by construction it is not something a search engine holds about you. It is genuinely useful for ranking two sites against each other on one scale on one day, and we use it for exactly that. What it is not is a defect, because there is nothing to fix, only a number that moves when somebody else’s crawler finds something.
Markup validation for its own sake. A validator will find dozens of issues on pages that render perfectly and rank fine. Fix what breaks rendering. The rest is a report with no reader.
None of these survive question one. They survive on every other list because a long list looks more thorough than a short one, and because a scoring tool needs a supply of deductible items to produce a number out of, which is a different job from finding out what is wrong with your site. If you want the honest version of what those tools measure, we went through it in our comparison of the free SEO tools and what each one actually checks.
The checks you cannot run without a crawler
This is the part most lists omit, and omitting it is what makes them impossible to finish. Several genuinely important checks are site-wide by nature. You cannot establish that no page carries an accidental noindex by opening pages one at a time, because the one you did not open is the one that has it.
Being concrete about the constraint helps. Screaming Frog, the crawler most audits reach for, publishes a free version whose crawl limit is 500 URLs, with the limit lifted on a paid license. For a small brochure site, 500 URLs is the entire site and the free tier is genuinely enough. For anything with a blog archive of any size, a catalog, or faceted URLs, 500 is a sample and you need to be honest that it is one.
When you do not have a crawl, Search Console replaces a meaningful part of it and people underuse it. Its page indexing report tells you which URLs are not indexed and groups them by reason, which covers most of tier one across the whole site at once. Its URL inspection tool tells you what was actually retrieved and rendered for a specific page, which settles blocked-resource questions properly rather than by guessing. Its links report tells you which of your pages you link to most internally, which is the tier two nomination question answered from data instead of from memory.
What Search Console will not give you is titles and canonicals across every page in one table, redirect chains, or an inventory of what links to what. Those need a crawl. Say so in the audit rather than quietly skipping them, because an audit that does not state its own coverage is not a finding, it is an impression.
How long the list and the audit should be
A full SEO website audit checklist that fits this shape runs to roughly two dozen items, not two hundred. The tier one checks take ten minutes. Tier two takes an hour or two of real thought and is where the value is. Tier three takes an afternoon and then does not need touching for a long time.
The pressure to pad comes from the checklist competing with other checklists rather than serving the person running it. Resist it by asking, of any item you are tempted to add, what you would tell someone to do if it failed. If there is no answer, or the answer is that it probably does not matter much, that is the item telling you where it belongs.
Be equally careful about what you promise afterward. Google’s own guidance on timing is that some changes might take effect in a few hours while others could take several months, and that in general you likely want to wait a few weeks before assessing whether your work had a beneficial effect. A checklist that implies you fix things on Monday and read the result on Friday is setting up a false verdict on real work.
The full SEO website audit checklist in one place
Tier one, stop if any of these fail. Robots file permits what you want crawled. No unintended noindex on key templates. Canonicals point at the page itself. Key templates return a success status. Rendering resources are not blocked.
Tier two, where the work is. One clear page per query you care about. The queries a page receives match the query you built it for. Internal links nominate the page you intended. Titles describe the page, not the site. Overlapping pages identified and given a verdict. Each important page answers the query in its opening rather than further down. Pages a visitor needs are reachable from navigation rather than only from a sitemap.
Tier three, once and properly. Sitemap exists, is reachable, and lists only what you want indexed. Redirects resolve in one hop. Structured data matches the page. Hreflang reciprocates, if you use it. Structure left alone unless there is a reason beyond tidiness.
Coverage, stated honestly. Record whether you crawled the site or sampled it, how many URLs you saw, and which checks you could not complete. An audit with a stated boundary is worth more than one that implies it looked everywhere.
Not on the list. Keywords meta tag, keyword density targets, keywords in the domain, crawl budget on a small site, vendor authority scores, validator output that does not affect rendering.
If you would rather work from a document than a web page, we keep a free SEO template you can copy that carries the same structure in a form you can edit.
What this checklist does not cover
Stating the boundary is part of the artifact. This is a site-wide SEO checklist and several audits that share the word are genuinely different work.
A business with a physical location is graded on factors that are not on your website at all, which is why the local SEO audit checklist is a separate list built around the profile rather than the site. A brand with many locations under one template has a duplication problem with a shape all its own, handled in the franchise website audit checklist. A catalog introduces faceted URLs and out-of-stock handling that a brochure site never meets, which we cover in the ecommerce site audit. At genuine scale the constraint stops being knowledge and starts being coordination, which is the subject of enterprise technical SEO auditing.
Two more boundaries worth naming. This list does not tell you how to write up what you find, which matters more than people expect, since a finding nobody acts on was not worth discovering. And it does not argue the case for auditing at all, which we made separately in why an SEO audit is worth running.
What we would check first, and what to ask us
Given one hour on a site we had never seen, we would spend ten minutes on tier one, twenty minutes in Search Console’s page indexing report finding out what is excluded and why, and the remaining half hour on the question of which page is competing for the queries that matter. That is the honest allocation, and it is unglamorous, because most of what an audit finds on a healthy site is that the site is fine and one thing is wrong.
If you are commissioning an audit rather than running one, the question worth asking is what the auditor did not check and why. Anyone who has actually done the work can answer immediately, because they made those decisions consciously. An audit that claims complete coverage of a site of any size has either not stated its sampling or has not thought about it. You can see how we handle coverage and sampling on our SEO audit page.
