SEO tasks: 12 jobs sorted by how often they need doing
Weekly, monthly, quarterly and yearly SEO tasks, sorted by how fast each thing breaks, including the 16-month Search Console deadline nobody is reminded of.
On 27 August 2026 this site replaced its portrait photo, deployed correctly, and kept serving the old one. The origin server returned the new file, 49,051 bytes; the domain, through its cache, returned the old one, 88,111 bytes. Every check that looked at the deployment said the job was done. The only check that would have caught it was opening the page the way a visitor does. Most recurring SEO lists are built around reports, and the failures that cost the most never reach a report. This list is sorted by how fast the thing each task checks can change, from weekly to yearly, and it ends with two tasks that do more harm as a routine than not doing them at all.
What SEO tasks should be done regularly?
Regular SEO tasks are sorted by how fast what they check can change: weekly for indexing errors and the live site, monthly for Search Console queries, yearly for exporting data before Search Console's 16-month window drops it. A new site does less; a mid-authority site adds content reviews; a large one audits templates.
Quick answer
- Weekly tasks catch silent breakage: the live page, the indexing report, the files that control crawling.
- Monthly tasks read demand: queries against a fixed set, pages with impressions and no clicks, links to new pages.
- Yearly tasks protect the data itself. Search Console keeps 16 months, and nothing older comes back.
- Some tasks should never be routine. Google says repeated indexing requests for the same URL do not speed up crawling.
| Frequency | Task | What breaks if it is skipped |
|---|---|---|
| Weekly | Open the pages that changed, on the domain | A visitor sees something no report shows |
| Weekly | Read the page indexing report | Pages drop out and nobody notices for months |
| Weekly | Re-read robots.txt and the sitemap after any release | One leftover line blocks the whole site |
| Monthly | Compare a fixed set of queries | An average moves and gets misread as a trend |
| Monthly | Find pages with impressions and no clicks | Positions are earned and wasted on titles |
| Monthly | Link new pages from older ones | New pages sit orphaned |
| Monthly | Glance at Core Web Vitals field data | A template regression spreads to every page |
| Quarterly | Re-open the sources cited on key pages | A quoted guideline changes and the page keeps saying the old one |
| Quarterly | Search the brand name | Wrong information about the business spreads |
| Quarterly | Review what was published and never shown | Thin pages accumulate |
| Yearly | Export Search Console data | Month 17 disappears |
| Yearly | Check the GA4 data retention setting | Explorations lose their history |
What kind of problem this is
The usual model of SEO maintenance is a monthly report: rankings, traffic, a list of errors from a crawler, and a few recommendations. It treats every task as having the same rhythm, and it is designed around what tools can export.
Two things are wrong with it. The first is that the fastest failures are not in any export. A cached page, a deploy that did not reach the domain, a noindex left over from staging: these break in hours and show up in a monthly report only as a traffic line that dipped for reasons nobody can name. The second is that the slowest risks have deadlines tools do not announce. Search Console does not warn anyone before the oldest month of data falls out of the window.
Sorting tasks by how fast their object changes fixes both problems at once. It puts the cheap, frequent checks where breakage is fast, and it puts a date on the tasks that protect data before it is gone. It also sits on top of the setup work, which is a different job: the one-time items, in order of return, are in the basic SEO checklist.
Weekly: the things that break silently
1. Open the pages that changed, on the domain
The check that matters is the page a visitor receives, not the content management system, not the deploy log and not the origin server. The case at the top of this article is the reason. The verification that was done tested what had been deployed, which was correct, and not what the domain was serving, which was a cached copy of the old file. Only the second tells you what a visitor sees.
The task takes minutes: every page that changed this week, opened from the public address, in a browser that is not logged in. For a site with no weekly changes, this task is empty, and that is fine.
2. Read the page indexing report
Search Console's page indexing report shows which pages Google has indexed and why the rest are not. A weekly look is enough to catch a jump after a release, which is when most indexing problems are introduced. The diagnosis for each status, including pages that are crawled and deliberately not indexed, is in why a page is not indexed.
3. Re-read robots.txt and the sitemap after any release
A Disallow: / carried over from a staging site takes one line to cause and removes an entire site from crawling. The task is only needed in weeks when something was released, and in those weeks it is the most valuable minute on the list. The broader set of checks for a release, before and after it goes live, is in the website launch checklist.
Monthly: reading demand without fooling yourself
4. Compare a fixed set of queries, not the average
An average position is a mixture of queries, and the mixture changes every month. When a site starts appearing for new, deeper queries, its average position gets worse while nothing has been lost. This site's own Search Console for August 2026, consulted on 5 September 2026, shows the effect: an average position of 72.2 with 92.3 % of impressions beyond position 30. That average is a measure of how much room is left, not a performance score.
The monthly task that avoids the misreading is a fixed cohort: the same set of queries, taken from a complete month, compared month to month. On 5 September 2026, the 13 top queries of August showed no drop of five positions or more, which meant the positions held. How to read a real drop against a mixture effect is in why organic traffic dropped.
5. Find pages with impressions and no clicks
Search Console lists pages and queries that appear in results and are not chosen. Those are usually title problems rather than content problems, and they are the cheapest improvement available on most sites. A monthly pass through the queries report sorted by impressions finds them.
6. Link new pages from older ones
A page that nothing links to is harder for search engines and readers to reach, and publishing steadily produces them without anyone deciding to. In one audit of a client's recipe site, 35 pages had no internal links at all, and together they carried roughly 180,000 impressions a quarter that the site's own structure never pointed to. The monthly task is to check that every page published that month has at least one link from an older page on the same topic.
7. Glance at Core Web Vitals field data
The Core Web Vitals report in Search Console groups URLs that share a similar experience, which is why a monthly glance is enough: a problem usually comes from a template and spreads to every page built on it. What the report measures and where it stops mattering is in Core Web Vitals in Search Console.
Quarterly: the things that age
8. Re-open the sources cited on key pages
Documentation from Google and other third parties changes without notice, and a page that quotes it keeps quoting the old version. This project met a clear example: a timeframe of four months to a year, widely attributed to Google's documentation, was no longer in it when checked on 12 August 2026. What the guide says now is that some changes take effect in hours and others take several months, and that waiting a few weeks before assessing the effect is sensible.
The quarterly task is to open every external source cited on the pages that bring the most traffic and confirm that each still says what the page says it says.
9. Search the brand name
What appears for the business name, including a Business Profile, directory listings and review sites, is often the first thing a new customer checks. A quarterly search catches outdated hours, a closed location still listed, or a directory with the wrong phone number. For local businesses the recurring layer is longer and has its own rhythm, set out in the local SEO checklist.
10. Review what was published and never shown
Pages that have existed for a year and never received an impression in Search Console are candidates for merging, improving or removing. A quarterly review keeps them from piling up. On a small site there may be none, and that is a good result, not a skipped task.
Yearly: protecting the data itself
11. Export Search Console data
Search Console's performance data covers the last 16 months. Google's own guide to debugging traffic drops puts the workaround plainly: "If you'd like to extend the 16 months, you could use the Search Analytics API or bulk data exports to pull data and store it in your systems."
This is the task with the hardest deadline on the list, and no tool reminds anyone of it. A yearly export keeps a four-month overlap with the live window, which is enough to compare the same month across years once the site has history. How to set up Search Console so the data starts accumulating from the first day is in how to set up Search Console.
12. Check the GA4 data retention setting
Google Analytics 4 standard properties offer two retention settings, 2 months and 14 months, and Google's help specifies that "the data retention setting only affects explorations and funnel reports." Standard reports are not affected. A property left on 2 months can only explore the last two months of that data, and nothing in the interface flags it. The task is a yearly confirmation that the setting is still what it should be.
Two tasks that should never be routine
Requesting indexing, again and again
Google's page on asking for a recrawl is direct: "Crawling can take anywhere from a few days to a few weeks", "Requesting a crawl does not guarantee that inclusion in search results will happen instantly or even at all", and "requesting a recrawl multiple times for the same URL won't get it crawled any faster." A weekly habit of resubmitting pages uses a quota and changes nothing.
Checking rankings every day
Positions move from day to day for reasons that have nothing to do with the site, and Google's guide says some changes take several months to show. A daily check produces decisions from noise. The honest timeframe, and why nobody can promise one, is in how long SEO takes.
When NOT to follow this list
When the site is new and has almost no data. Tasks 4, 5 and 10 need months of impressions to mean anything, and doing them early produces conclusions from a handful of clicks.
When nothing changed. The weekly tasks exist because releases break things; a site that was not touched this week has no weekly work.
When the list becomes the job. Maintenance keeps a site from losing ground, and it does not create demand. A monthly report that shows these twelve tasks completed says nothing about whether the site is growing, which is the question behind how to tell if your SEO agency is working.
What the first results leave out
Search for this topic and most results are monthly checklists. SEO Buddy's guide to monthly SEO tasks, published on 18 August 2026, is one of the better ones: five tasks with time estimates of 60, 120, 60, 45 and 90 minutes, and a useful warning, quoted from Google, against requesting indexing as a routine. It does not mention the 16-month limit on Search Console data, the GA4 retention setting, or checking the live page rather than the report.
The pattern is consistent across the first page, checked on 14 September 2026. Lists organized by month put everything on the same rhythm, so the fastest risks get checked too rarely and the data deadlines never get a date at all. In Ahrefs the same day, "seo tasks" is searched about 900 times a month in the United States with a keyword difficulty of 5, while "seo to do list", at about 350, has "seo checklist" as its parent topic: people searching for recurring tasks and people searching for a setup checklist are treated as two different questions, which is how this list and the checklist are split.
Mistakes that repeat
- Verifying the deploy instead of the domain. The origin can be right while visitors receive a cached copy.
- Reading average position as performance. The average moves with the mix of queries.
- Waiting for a report to show a robots.txt error. By the time it shows, the damage is weeks old.
- Never exporting Search Console data. Month 17 is gone for good.
- Leaving GA4 retention at 2 months. Explorations lose their history without warning.
- Resubmitting pages for indexing every week. Google says it does not make crawling faster.
Data and transparency
The portrait case, with the origin returning 49,051 bytes and the domain returning 88,111 through its cache on 27 August 2026, is this site's own measurement. The August 2026 figures, an average position of 72.2 with 92.3 % of impressions beyond position 30, and the cohort of 13 queries with no drop of five positions or more, come from this site's Search Console, consulted on 5 September 2026 for the window of 1 August to 3 September 2026. The 35 orphan pages and roughly 180,000 impressions a quarter come from an audit of a client site, which is not named, with the figures rounded.
That Search Console performance data can be viewed for the last 16 months and extended with the Search Analytics API or bulk data exports comes from Google's guide to debugging drops in Search traffic; the GA4 retention options and what they affect from Google Analytics help; and the three statements about recrawling from Google's page on asking Google to recrawl, all opened on 14 September 2026. That a four-month to one-year timeframe no longer appears in Google's guide, and what the guide says instead, was checked on 12 August 2026. SEO Buddy's monthly tasks guide, published on 18 August 2026, was read on 14 September 2026. Search demand for "seo tasks" (about 900 a month, KD 5), "seo to do list" (about 350, KD 41, parent topic "seo checklist") and "monthly seo tasks" (about 50, KD 0) comes from Ahrefs for the United States, pulled the same day. The frequencies assigned here are operating judgment from audit work across a portfolio recording more than 300 million impressions a year in Search Console. Verified as of September 2026.
Primary sources, opened on 14 September 2026: debugging drops in Google Search traffic; GA4 data retention; ask Google to recrawl your URLs.
What this changes
Maintenance lists are usually judged by how complete they are. The ones that fail most often are complete: every task is on them, all on the same monthly rhythm, and the report that records their completion looks the same in a good month and a bad one.
The useful question about a task is not whether it is on the list but how long its object takes to break. A cached page breaks in an hour and a data window closes after sixteen months, and a monthly rhythm is too slow for the first and gives no date to the second. The two most expensive failures a site can have are a page nobody sees correctly and a history nobody can get back, and neither one ever shows up as a line in a report.
Frequently asked questions
What SEO tasks should be done every week?
Three short checks, and only when something changed: opening the updated pages on the public domain as a visitor would, reading Search Console's page indexing report for new exclusions, and re-reading robots.txt and the sitemap after any release. These catch failures that break in hours, such as a cached old page or a leftover staging rule, long before a monthly report would show them.
What SEO tasks should be done every month?
Comparing a fixed set of queries rather than the average position, finding pages with impressions and no clicks, linking pages published that month from older ones, and glancing at Core Web Vitals field data. The fixed set matters because an average position moves with the mix of queries, so it can worsen while a site is actually growing into new searches.
How long does Search Console keep data?
Search Console's performance data covers the last 16 months. Google's guide to debugging traffic drops says the window can be extended by pulling data through the Search Analytics API or bulk data exports and storing it elsewhere. A yearly export keeps an overlap with the live window, which preserves year-over-year comparisons once the site has more than 16 months of history.
Should I request indexing every time I update a page?
Occasionally, for an important change, and never as a routine. Google's documentation says crawling can take from a few days to a few weeks, that a request does not guarantee inclusion, and that requesting a recrawl multiple times for the same URL will not get it crawled any faster. Individual URL requests are also subject to a quota.
How many hours a month does SEO maintenance take?
For a small site with few changes, a couple of hours a month covers the weekly and monthly tasks here, and the quarterly and yearly ones add a few hours a year. SEO Buddy's monthly plan, published in August 2026, adds up to 375 minutes, but it includes content and authority work. The weekly checks take minutes and only exist in weeks with a release.
Most sites do not have a ranking problem
They have a what-happens-next problem. You can rank first and still sell nothing. The diagnostic looks at both and tells you which one is costing you money.
See the diagnostic