The SEO checklist: every check that matters, in order.

1. How to use this checklist.

Work the groups in the order they appear. The sequence is the point. Setup checks run once per domain and are pure disqualifiers, so they go first. Keyword checks run before a word is written, because everything downstream is applied to a page that already exists and cannot be un-chosen cheaply. On-page, links, and the technical floor run per page. Measurement runs on a clock you do not get to shorten.

Two thirds of the items carry an AGENT tag and one third a YOU tag. That split is not a pitch, it is the honest line between decisions and repetition. Whether a query is worth wanting, which siblings a page should link to, whether the business case for a topic exists: those stay with a person who knows the business. Validating volume, structuring the answer, writing the title, attaching sources, shipping schema, and wiring the links are mechanical, and doing them by hand 40 times a month is how programs quietly stop in month two. The whole method behind those per-page decisions is written up separately in our guide to making a blog rank, which is the method this page turns into a list.

One boundary, because the two get conflated. This is the whole-program checklist and it treats the technical layer as a floor: seven checks, each of which can cost you every position on a URL, none of which earn one. The line-by-line technical audit is a different job at a different depth, covering crawl budget, index bloat, redirect chains, log files, hreflang, and vitals diagnostics, and it gets its own checklist rather than a section here. If you already suspect the plumbing is the problem rather than the program, start with a crawl of your own site through the free SEO audit and come back to this list afterwards.

Last framing note before the checks. Most of this list is not a ranking lever. The genuine levers are the query you pick, how directly you answer it, how the page is structured, and what links to it. Everything else is the floor you have to be standing on for the levers to do anything, which is why a site can pass 37 of 40 checks and still get nothing: the three it failed were the ones that mattered.

2. The SEO checklist for a new website.

Eight checks, run once, before any content work. None of them will earn a position. Each of them can quietly prevent every position that follows, which is the worst failure mode in SEO: everything looks correct, the work continues for months, and the reason nothing moves is one line in a file nobody opened. Clear these on day one and you never have to wonder again.

1. 01YOU

   One canonical hostname, everything else redirected to it.

   Pick www or bare domain, http or https, trailing slash or not, then 301 every other variant to the one you picked. Four live variants of the same page is the most common way a new site splits its own signal.

2. 02YOU

   HTTPS on every URL, with a certificate that is not about to expire.

   Mixed content and expired certificates are both hard stops. Check the renewal is automated rather than a calendar reminder somebody will miss.

3. 03YOU

   Search Console verified, sitemap submitted.

   Without it you are guessing at indexation and cannot see a single query you rank for. It takes five minutes and it is the only free source of your own performance data.

4. 04YOU

   An XML sitemap that lists only indexable, canonical URLs.

   Staging routes, paginated duplicates, and noindex pages in the sitemap are a contradiction: you are asking for a crawl of a page you told search engines to ignore.

5. 05YOU

   robots.txt blocks nothing you want indexed.

   Read it line by line rather than assuming. A disallow left over from the build phase is the single most expensive one-line bug in SEO.

6. 06YOU

   A self-referencing canonical tag on every page.

   It costs nothing, it survives tracking parameters and syndication, and it removes the ambiguity that produces duplicate-content problems later.

7. 07YOU

   The blog URL pattern decided before the first post.

   Flat and short beats nested and dated. Changing it later means redirecting every post you published, so make the decision once and stop revisiting it.

8. 08YOU

   Analytics installed and confirmed recording a real session.

   Install it, load the site from a phone on mobile data, and watch the session land. An unverified tag is the reason so many sites have a six-month hole in their history.

A new site has one genuine advantage here: nothing has been built on top of the mistakes yet. Changing a URL pattern on day one costs nothing and on day four hundred costs a redirect map plus some fraction of the signal every one of those pages had accumulated. The same applies to the hostname decision and the blog structure. Make all three deliberately now, write them down, and stop revisiting them.

3. Keyword selection: six checks before you write.

This group decides whether the other 34 checks are worth running. A well-built page aimed at a query your site cannot win, or a query nobody types, is a well-built page nobody reads. Ahrefs studied roughly a billion pages and found 96.55% of them get zero organic search traffic from Google. Not a trickle. None. Target selection is the largest single contributor to that number, and it is also the cheapest thing on this list to get right.

1. 09AGENT

   The query has measured search volume, not assumed volume.

   Guessing what people type is the most expensive habit in SEO because the cost is invisible: the post ships, gets no impressions, and you conclude search does not work.

2. 10AGENT

   The current results page is winnable at your authority.

   Open the results and look at who holds the top five. If it is five sites with a decade of links at that phrase, pick the more specific question underneath it.

3. 11AGENT

   Intent read off the results, not off the phrase.

   If the top ten are all comparison lists, a how-to guide will not rank there no matter how good it is. The results page tells you what format is allowed to win.

4. 12YOU

   One sentence connecting the searcher to something you sell.

   If you cannot write it, the page is a brand-awareness play. That is a legitimate choice, but it should be a deliberate one rather than a discovery you make in month six.

5. 13AGENT

   No existing page on the site already targets this query.

   Two pages aimed at one query split the signal and neither wins. Check the archive before writing, and strengthen the existing page instead when there is a match.

6. 14AGENT

   One query assigned to one page, written down.

   A keyword-to-URL map is the cheapest cannibalisation defence there is. Without it, the fourth writer on the team re-targets a query you already own.

Two of these deserve more than a line. Winnable means specific: a six-month-old site going after a two-word category term is competing against a decade of accumulated links at that exact phrase, and the specific question underneath it is both winnable now and brings the variations with it. And worth wanting is the check people skip in the excitement of finding something easy, because traffic that never converts costs exactly as much to produce as traffic that does. Validation against live search data is the first stage of our SEO automation pipeline for that reason, and it kills more proposed articles than it approves. Fanning an approved topic out into the adjacent questions worth covering is what the blog post idea generator is for.

4. The SEO optimization checklist, per page.

Nine checks, applied to every page, every time. The highest-leverage one is the first: move the answer to the top. Most pages open with context, a paragraph of throat-clearing, and a promise that the answer is coming. Delete all of it. Put two to four sentences that fully resolve the query inside the first hundred words, complete enough that they still make sense lifted out of the page.

1. 15AGENT

   The query answered inside the first hundred words.

   Two to four self-contained sentences that resolve the question before any preamble, complete enough to still make sense cut out of the page.

2. 16AGENT

   Exactly one H1, carrying the target phrase.

   Templates that render the logo or the site name as an H1 produce two. Check the rendered page rather than the editor.

3. 17AGENT

   One H2 per follow-up question a reader would ask next.

   Each section answers its own heading and stands alone. This is what makes a single page eligible for dozens of related queries it never explicitly targeted.

4. 18AGENT

   A unique title tag that leads with the query.

   Roughly 60 characters before truncation. The title is the click decision: impressions with no clicks is almost always a title problem rather than a content problem.

5. 19AGENT

   A meta description of 110 to 160 characters, written for the click.

   It is not a ranking factor and it does change click-through. Leave the field empty and search engines write one for you out of whatever text is nearest.

6. 20AGENT

   A short lowercase slug containing the query, set once.

   No dates, no stop words, no category prefixes you might reorganise. Changing a slug later costs you a redirect and some of the signal it had accumulated.

7. 21AGENT

   Every specific claim attached to a named source.

   A sourced figure is quotable by a reader, a journalist, and an answer engine. A confident adjective is quotable by nobody.

8. 22AGENT

   Images compressed, with alt text that describes the image.

   Alt text describes what is in the frame. Stuffing the target keyword into it helps nothing and makes the page worse for anybody using a screen reader.

9. 23AGENT

   Article schema on the page, with headline, author, and dates.

   It does not lift rankings. It removes ambiguity and makes the page eligible for richer presentation, and it costs ten minutes once per template.

Notice what is not on this list: a word count, a keyword density figure, and a target number of headings. Google Search Central's guidance on creating helpful, reliable, people-first content names no preferred length anywhere and asks instead whether the page provides substantial value compared with what already ranks. Length is a symptom of a complete answer, not a target to hit, and padding a finished 900-word answer out to 2,000 words dilutes the passages that were working.

Three of these nine have tools attached because they are the ones people rush. The title is the click decision, which is what our headline generator exists to fix. The description will be rewritten for you if you leave it empty, so write it with the meta description tool instead. And every specific claim needs a named source behind it, which is the standard our fact-checking pass enforces on every draft before it can reach the publish step.

Schema is on the list but belongs in perspective. Google's documentation on Article structured data describes it as helping Google understand the page and making it eligible for richer presentation, not as a ranking factor. Ship it anyway, because it removes ambiguity about headline, author, and publish date at a cost of about ten minutes once per template. Just do not expect it to move anything on its own.

5. Internal links and site structure.

Five checks, run at publish time rather than in a quarterly audit, because publish time is the only moment anybody has the context to do them well. Almost every site links in one direction only: the new page points back at older pages, and nothing is ever edited to point forward at the new one. The archive ends up hoarding all the internal signal while every new page starts from nothing, which is exactly backwards from what you want.

1. 24AGENT

   The new page links up to the pillar page for its topic.

   One deliberate link from the spoke to the hub, with an anchor that reads as the phrase the hub is meant to rank for.

2. 25AGENT

   It links sideways to two or three siblings.

   Siblings are the pages covering adjacent questions. These are the links readers actually follow, which is why they are worth choosing rather than automating blindly.

3. 26AGENT

   At least two older pages edited to link down to it.

   Almost every site links in one direction only, so the oldest posts hoard the internal signal and every new page starts cold. This check is the fix.

4. 27AGENT

   Anchor text varied per target across the page.

   A page effectively counts the first anchor per target, so repeating the identical phrase four times burns three slots. Write natural variations instead.

5. 28YOU

   No orphans: every URL reachable from another page in three clicks.

   A page nothing links to is discovered late, valued little, and usually forgotten by the person who wrote it within a week.

Anchor text is the part worth being precise about. Write the anchor as the phrase you want the destination to rank for, inside a sentence that reads naturally without it. Click here is a wasted slot, and so is pointing at the same page from eleven posts with eleven identical anchors. If you want a read on where an existing archive is leaking, our internal linking tool maps what points at what. Do this consistently and the archive becomes a structure rather than a list, which is the same mechanism behind building topical clusters at scale: a hub holding the head term, spokes each owning one sub-question, links running both ways.

6. The technical floor.

Seven checks. Short, boring, and disqualifying. Nothing here wins a position and every item can cost you one, which is why it sits this late in the list and still has to be cleared. Run the whole group as a crawl of your own site once a quarter rather than as a vague intention to look into performance at some point.

1. 29YOU

   The page is actually indexed, confirmed rather than assumed.

   Check coverage in Search Console. Published, live, and in the sitemap are three different things from indexed, and only the last one can rank.

2. 30YOU

   No stray noindex or nofollow left on a template.

   These get added during a staging build and never removed. Check the rendered head of a real production URL, not the template source.

3. 31YOU

   Core Web Vitals inside the thresholds on real traffic.

   Largest contentful paint under 2.5 seconds, interaction to next paint under 200 milliseconds, cumulative layout shift under 0.1, measured on field data rather than a lab run.

4. 32YOU

   The mobile page carries the same content as the desktop page.

   Indexing is mobile-first. Copy hidden or dropped from the small-screen layout is copy the index may never see at all.

5. 33YOU

   The content is in the HTML, not only injected after load.

   View source and search for a sentence from the middle of the page. If it is not there, you are betting the whole page on client-side rendering being crawled well.

6. 34YOU

   No duplicate titles or meta descriptions across the archive.

   Duplicates cluster in paginated archives, tag pages, and anything templated. Crawl your own site quarterly and fix whatever the crawl finds.

7. 35AGENT

   A publishing path that needs no manual reformatting.

   If getting a finished page live means re-uploading images and fixing the schema by hand, the cadence dies on that step rather than on the writing.

The vitals thresholds are specific, so use the real numbers rather than a general sense that the site feels fast: Core Web Vitals sets largest contentful paint at 2.5 seconds, interaction to next paint at 200 milliseconds, and cumulative layout shift at 0.1, measured at the 75th percentile of real visits rather than on your laptop. A lab score of 98 on a wired connection tells you almost nothing about what a phone on mobile data experiences.

The last check in this group is the one that looks least technical and breaks the most programs. If getting a finished page live involves reformatting the markup, re-uploading the images, and fixing the schema by hand, the cadence dies at that step. Not at the writing, not at the research, at the twenty minutes of manual labour standing between a finished draft and a published URL. Native publishing straight into the CMS, which is what our WordPress publishing integration does, exists because that friction is where most content programs actually stop.

7. Links and measurement.

Five checks that close the loop. Links first, and the useful fact here is how low the bar is: the same Ahrefs study found 66.31% of pages have no backlinks pointing at them at all. You are not competing against a web where every page is well linked. You are competing against a handful of pages per query that have any links, which is why pointing the few you can earn at the pages targeting commercial queries beats funnelling all of them at the homepage.

1. 36YOU

   The obvious profiles and listings that name the site are claimed.

   Not a link-building strategy. It is the baseline set of places an entity is expected to exist, and it is a one-afternoon job you do once.

2. 37YOU

   Links point at the pages you want to rank, not only the homepage.

   Most link building funnels everything to the homepage, then the pages that actually target commercial queries stay at zero.

3. 38YOU

   Impressions checked at 30 days, not clicks.

   Zero impressions after a month is a discovery or a targeting problem and it is fixable now. Zero clicks at a month means nothing at all yet.

4. 39YOU

   At 90 days, position read before clicks.

   Impressions with no clicks means the page ranks on page two or three and the snippet is not earning the click: rewrite the title before touching the body.

5. 40YOU

   A quarterly pass over the top twenty pages.

   Update the figures, replace the dead sources, add links to whatever you published since. It is consistently cheaper per position gained than a new page.

The measurement checks are on a clock you do not control. Ahrefs tracked newly published pages and found only 5.7% of them reached a Google top 10 for any keyword within a year, with the successful ones mostly taking two to six months. Judging a page at week four and rewriting it destroys the only signal you had and restarts the clock, which is how a site spends a year permanently at day one.

One scorecard adjustment. Sessions alone stopped being a fair read once AI answers landed in the results: Pew Research Center instrumented US adults and found people clicked a traditional result on 8% of visits where an AI summary appeared against 15% where none did. Fewer clicks per query, arriving later in the buying process. The page structure that earns a citation in one of those answers is the same structure in section four, and the extra work specific to answer surfaces is covered in our answer engine optimization guide and built into every draft by the AI search optimization layer.

A worked example: one page, checked.

Take a fictional field-service scheduling product, Dispatchly, with a page targeting how to schedule HVAC technicians. It is 1,900 words, well written, live for seven months, sitting at 60 impressions and no clicks. Here is what running the checklist over it actually returns. The company and its numbers are invented. The distribution of failures is not: this is the shape almost every audit of a well-written page that is not working comes back in.

The distribution is the lesson. Thirty-one passes and the page earns nothing, because the nine failures were concentrated in the two groups that hold the actual levers: what the page is for, and whether anything points at it. This is why running a technical audit on a page that is not working so often returns a clean report and no answer. The technical layer was never the problem.

Run this pass over the ten or twenty existing pages that carry real commercial intent before you publish anything new. For most sites it is a bigger win than the next three months of fresh output, and it takes about two hours per page with no writing involved at all.

Common mistakes with SEO checklists.

1. Running the list out of order. The on-page group gets run first because it is the most satisfying, and the keyword group gets run never. Every check after the target selection is applied to a page that already exists, so a wrong target makes the other 39 items polish.
2. Treating disqualifiers as if they were levers. Shipping schema, adding a canonical, and compressing an image are all worth doing and none of them will move a position. Teams work the floor for a month, see nothing, and conclude search is rigged.
3. Auditing the technical layer first. It is the easiest layer to measure, so it absorbs the attention. On a page that is well built and still not ranking, that group is almost always already green and the failures are upstream in the target or downstream in the links.
4. Checking the list once, at launch. A checklist run once is a launch ritual. The keyword and on-page groups only work as a gate applied to every page, every time, which is precisely the part that erodes first when the team gets busy.
5. Confusing a list with a program. Ten pages that pass 40 checks each, all on unrelated topics, leaves you shallow in ten places. Coverage of one subject is what makes a site read as knowing the subject, and no per-page checklist captures that on its own.
6. Hitting the cadence by dropping checks. The rate is held, the standard slips, and a site full of thin pages is a worse signal than a site with three good ones. Either fix the production constraint or lower the rate honestly.
7. Judging the work at 30 days. Most pages that rank take two to six months. Rewriting at week four throws away the only data you had, and doing it repeatedly keeps the whole archive permanently at day one.