How many blog posts should you publish a week?

For most small teams the answer is one. Make it two if both can be researched, checked and edited properly by the people you actually have. Fewer than one is fine if each piece is the best available answer to a question your customers ask. Anything above two a week is a staffing decision first and a content decision second.

That is the short version. The longer version matters because the number you pick quietly decides everything downstream: how deep each article goes, whether anyone checks the facts, and whether the blog grows into a library or a pile.

What Google says about frequency

Google does not reward you for posting often. Its Search Central documentation on creating helpful, reliable, people-first content says its automated systems aim to reward useful, original content made for people, as opposed to content made to manipulate rankings. Google's search advocates have said publicly that publishing frequency or volume on its own is not a ranking factor.

The guidance has since been sharpened. Google's October 2026 revision names four qualities its quality raters look for in a page's main content: effort, originality, talent or skill, and accuracy. It also warns against two habits that cadence-chasing tends to breed. One is publishing thin content at volume. The other is changing dates to look fresh when nothing of substance has changed.

Freshness does count for some searches. Google applies freshness systems to queries where people expect recent information, such as breaking news, trending topics and recurring events. If you sell accounting software or run a landscaping firm, most of your subjects are not like that. For evergreen and commercial topics, what matters is whether the page is accurate and still relevant. A two-year-old setup guide that describes buttons your product no longer has fails that test, however busy the blog around it looks.

Crawled, indexed and cited are different things

Publishing more does change one thing: how often crawlers visit. Google's crawl budget guide explains that crawling depends on how much load your server can take and how much demand Google has for your URLs. Frequent publishing on large or fast-changing sites can raise that demand. The same guide is plain that a crawl guarantees neither indexing nor ranking.

It helps to keep three words apart. Crawled means a bot fetched the URL. Indexed means the page was parsed and stored. Ranked or cited means it was picked as an answer to a query. Volume mostly affects the first of these. The other two depend on what is on the page.

AI assistants follow the same logic. OpenAI documents separate agents for separate jobs. OAI-SearchBot finds and indexes pages for ChatGPT's search features. GPTBot collects content for training models and is controlled separately in robots.txt. ChatGPT-User fetches a page when a person's prompt calls for it. Block OAI-SearchBot and your pages cannot appear in ChatGPT search answers. Allow it, and OpenAI still says that does not guarantee a citation in any particular response.

Perplexity documents PerplexityBot, which indexes pages for discovery and citation in its results, and Perplexity-User, which handles live fetches during a search. Its retrieval works on passages inside a page, so an answer may lift a single passage from your article. That passage has to make sense on its own and be correct. The effort belongs in each article, and a fourth thin post on the same subject adds nothing to it.

The tradeoffs at each pace

If you are choosing between paces, compare them on the same terms.

The costs in the bottom row tend to surface in Search Console. When a site publishes faster than it can make pages worth indexing, URLs collect under "Discovered - currently not indexed" in the Page indexing report. Google found them. It has not yet decided they are worth storing.

Let demand set the number

A fixed quota sounds disciplined. Three posts a week, every week, looks good on a planning board. In practice it forces you to publish in weeks when you have nothing worth saying, and the pages written to fill a slot are the thin ones Google's guidance describes.

Work the other way round. Gather the questions customers actually ask: sales calls, support tickets, the queries already showing in Search Console, keyword data for your category. Keep only the topics that pass two tests. Someone searches for it or asks it, and you can answer it better than the pages currently answering it. Then do the arithmetic. Thirty strong topics you want covered in six months comes to a little over one article a week. Twelve strong topics over the same six months makes you a twice-a-month publication, which is a perfectly respectable thing to be.

Repeat the exercise each quarter. The list grows as your product, your customers and your search data change, and the cadence should move with it.

New sites and mature sites need different weeks

A new publication is mostly gaps. Nobody can find you for the basic questions in your field because you have not answered them yet. For the first few months a faster pace can make sense, so the foundational articles in each topic cluster exist and Search Console starts returning data you can learn from. Two a week is reasonable here if each one is still checked to the same standard as the first.

A blog with a hundred articles has a different job. Much of its week should go to maintenance: updating articles whose facts have aged, merging posts that cover the same question, and retiring pages nobody needs. Count a substantial update as part of your cadence. Do not change the publication date on an article you have barely touched; Google's guidance names that practice specifically.

What your team can sustain

A long-form article done properly involves research, drafting, verifying every source, images and the fiddly last mile of the CMS. For one or two people that adds up quickly, and under deadline pressure the checking is the easiest step to drop. Choose the pace at which the checking still happens. That is your real ceiling, whatever the planning board says.

AI tools can raise the ceiling without lowering the standard, if you use them for the right part of the job. Google's guidance does not forbid AI-assisted drafting. It judges the result on helpfulness, accuracy and originality, which in practice means a person reviews the facts, the voice and the claims before anything goes live. Let software handle structure, first drafts and the mechanical steps. Keep a human on the facts and the final edit.

Rune, the AI blog writer that publishes this journal, is built around that split. It plans topics from Search Console, keyword and competitor data, fetches every source before citing it and drops any that will not load, then publishes to a hosted magazine or a Webflow CMS after review or on a set cadence. If you want to see whether that changes your arithmetic, the free preview writes two illustrated, sourced articles from your site and a short brief in about 5 to 15 minutes.

How to tell whether your cadence is working

Give any new pace about a quarter before judging it, then look at a few specific things.

In Search Console's Page indexing report, compare indexed pages with those marked "Discovered - currently not indexed". If the second group grows every time you speed up, you are publishing faster than you are publishing well. Crawl stats, under Settings, show total crawl requests and response times, which tells you whether Google is visiting more often and whether your server is coping. Impressions across the cluster you have been writing show whether the coverage is being seen at all.

For AI assistants, watch referral traffic in your analytics. ChatGPT links often carry utm_source=chatgpt.com, and Perplexity visits arrive with its own referrer. If you can read server logs, check for successful requests from OAI-SearchBot and PerplexityBot on your most important pages. A visit is not a citation, but no visit means no chance of one.

Finally, track organic sessions on older articles. When a once-reliable post starts to slide, updating it is usually a better use of a week than writing something new.

So pick one good article a week as the default, and move from there for reasons you can name: a long list of strong topics, a new site with obvious gaps, or a second person who can check the work. Drop to fewer when the list runs thin or old articles need attention. Search engines and AI assistants make no promises about any pace, and no number will do the work of a useful page. What you can control is that everything you publish was worth the reader's time and was checked before it went out.