Almost every guide on this topic gives you the same advice for every page on your site. Keep paragraphs short. Use headings. Write in active voice. Add a call to action.
That advice is fine and it is also incomplete in a way that costs you. A homepage, a service page, a comparison page and a 3,000 word guide have different jobs, different lengths and different structures. Applying one set of rules to all of them produces a homepage that reads like a blog post and a blog post that reads like a brochure.
We read the pages currently ranking for this query before writing. Semrush gives 10 tips, Jimdo gives 11 rules, Grammarly gives 7. Between roughly 7,000 words, none of them changes its advice based on which page you are writing, and there is not a single table between them.
This page is organised the other way round: by page type first, then the rules that apply to it.
Start by naming the job of the page
Before you write a word, answer one question: what does a visitor need to be able to do after reading this page? Not what you want to tell them, what they need to be able to do.
There are only a handful of real answers. Understand what you sell. Decide whether you are the right fit. Compare you against an alternative. Learn how to do something. Get in touch.
A page with two jobs does both badly. If you cannot name a single job, that is usually a sign the page should be two pages.
The spec for each page type
This is the table the ranking guides do not have. Lengths are targets rather than limits, and every row assumes the page is doing one job well.
| Page type | Job | Target length | Opens with | Judged by |
|---|---|---|---|---|
| Homepage | Say what you do and for whom | 300 to 600 words | One sentence naming the product and the buyer | Whether a stranger can repeat what you do |
| Service or product page | Make the fit decision easy | 600 to 1,200 words | The specific problem it solves | Enquiries or signups from the page |
| Comparison page | Help a decided buyer choose | 1,200 to 2,500 words | The one line summary of who each option suits | Conversions from high intent traffic |
| Guide or how-to post | Teach the task completely | 1,500 to 3,500 words | The answer, before the background | Rankings and time on page |
| Listicle | Cover the field credibly | 2,000 to 4,000 words | Methodology and who each entry suits | Rankings and citations |
| About page | Establish who is behind this | 400 to 800 words | Why the company exists | Trust, hard to measure directly |
| Landing page for an ad | Convert one specific promise | 400 to 900 words | The exact promise from the ad | Conversion rate |
The number people get most wrong is the homepage. It wants to be short, and the instinct to explain everything turns it into a page nobody finishes.
Why generic length advice fails
Several guides suggest 300 to 700 words as a reasonable length for web content. That is roughly right for a homepage and roughly half of what a guide needs to rank for anything competitive.
Length is not a quality signal on its own. It is a consequence of covering the topic. A page that fully answers a simple question in 400 words is finished, and padding it to 1,500 makes it worse.
The useful test is coverage rather than count. Look at what the pages currently ranking for your target query actually cover, then cover that plus something they do not have. If that takes 900 words, write 900.
Research before drafting, not after
The single biggest difference between content that works and content that does not is whether the topic was chosen from data or from a meeting.
For any page meant to attract search traffic, start with the query. Find what people actually type, check the search volume, and look at what currently ranks. If the top ten results are all product pages and you are writing a blog post, you will not rank, and the format is the reason.
For pages not meant to attract search traffic, like a homepage or an about page, the research is different but still necessary. Read your own support tickets and sales call notes. The words customers use are almost never the words the marketing team uses. When the goal is circulation rather than ranking, the rules change again, as our guide to how to create shareable content explains.
Structure the page before you write sentences
Write the headings first, in order, and check that they tell the story on their own. If someone reads only your headings, they should get the gist and know where to stop.
Each heading should be a claim or a question, not a label. "Pricing" tells the reader nothing. "What it costs for a team of five" tells them whether to keep reading.
Put the answer first inside each section. The inverted pyramid advice that the ranking guides give is correct, and most writers apply it only to the opening paragraph rather than to every section.
Write the draft badly on purpose
Getting a complete bad draft down is faster than getting a good one down in pieces. Write to the outline without stopping to fix sentences, then edit.
The reason this works is that editing and generating are different modes, and switching between them every sentence is what makes drafting feel slow. A complete rough draft also shows you the structural problems that no amount of sentence polish will fix.
Expect the first draft to be 20 to 30 percent longer than the final. Cutting is where most of the quality comes from.
Edit in three passes
Do not try to fix everything at once. Three passes, each looking for one thing, is faster and catches more.
The first pass is structural. Does each section deliver what its heading promises, is anything missing, and is anything in the wrong order? Move and delete freely here.
The second pass is for clarity. Cut the words that are doing no work, break paragraphs longer than four sentences, and replace the vague claims with specific ones. "Significantly faster" becomes "about three times faster" or it comes out.
The third pass is proofreading, and it is the only pass where a tool helps much. Read it aloud or have something read it to you, because the ear catches what the eye skips.
Writing so an AI answer engine can quote you
This is the part missing from every guide ranking for this query, and it has changed what good web writing looks like.
A large and growing share of searches now end without a click. The page still gets read, by an answer engine that lifts a sentence or two and cites the source. Our own Search Console data shows query sets with 14,000 impressions at an average position of 8.5 and zero clicks.
Writing for that outcome is not a separate discipline, but it does reward specific habits. Answer the question in the first two sentences of the section, before any context. Make claims that stand alone without the paragraph around them. Use specific numbers rather than ranges. Define the term you are explaining in a single sentence somewhere on the page.
The test is simple. Take any sentence out of your page, put it on its own, and see whether it still says something true and useful. Sentences that only work in context are invisible to the systems now doing the quoting.
Optimising without writing for robots
Keyword stuffing has not worked for many years, but the opposite mistake is just as common: writing a page so carefully natural that it never states its subject plainly.
Use the target phrase in the title, in the first hundred words, and in at least one heading. Then write normally. If the phrase appears naturally after that, fine, and if it does not, do not force it.
Cover the related questions people ask about the topic, because that is what actually correlates with ranking now. The related searches at the bottom of a Google results page are the cheapest source of these.
Internal links are part of writing, not a cleanup task
Every new page you publish starts with no internal links pointing at it, which means it starts invisible to both crawlers and readers.
Add two or three links out to relevant existing pages while you draft, where a reader would genuinely want them. Then find two or three existing pages that should link to the new one, and add those links.
If you are writing a set of related pages rather than a single one, plan the linking as a group from the start. Our roundup of content hub examples shows how the stronger sites arrange a pillar page and its supporting posts.
The second half is the part everyone skips, and it is the half that matters more. A page nothing links to will underperform whatever its quality.
The pre-publish checklist
Run this before anything goes live. It takes five minutes and catches most of what goes wrong.
| Check | Fails when |
|---|---|
| The page has one job | You cannot state it in a sentence |
| Title matches what the page delivers | The best part is not in the title |
| Answer appears above the fold | The first screen is background |
| Headings tell the story alone | Headings are labels like "Overview" |
| No paragraph over four sentences | Any wall of text survives editing |
| Specific claims, not vague ones | "Significantly", "a lot", "many" appear unqualified |
| Two or three outbound internal links | Zero, or a list dumped at the foot |
| Inbound internal links added | Nothing on the site links to it |
| One clear next step | Two competing calls to action |
Common mistakes worth naming
Writing for your peers rather than your buyer. Industry vocabulary makes a page sound credible to people already in the industry and opaque to everyone else. The buyer is usually less expert than the writer.
Burying the point under context. Most drafts spend three paragraphs explaining why the topic matters before saying anything. Readers who searched for the topic already know why it matters.
Leaving the page with no next step. A page that informs and then stops is a dead end. The next step does not have to be a purchase; a link to the logical next page is often better.
Publishing and never revisiting. Content decays, and the highest return writing work on most sites is updating pages that already rank. Our guide to running a website content audit covers how to find them.
Tools, and how much they matter
Tools help at the edges. They do not fix a page with the wrong job or the wrong structure, which is what is usually wrong.
A grammar checker is worth having for the proofreading pass. A content optimisation tool like Surfer or Clearscope is useful if you write a lot of competitive search content and unnecessary if you do not. Search Console is the only one that is genuinely essential, and it is free.
Be careful with AI drafting tools. They are good at producing structurally correct text and bad at knowing which claims are true, so anything factual they produce needs checking before it goes near a live page. If you use them at scale, the failure modes are worth understanding first.
The same caution applies to writing different versions of a page for different audiences, which is easy to generate and hard to keep accurate. We covered where that works and where it backfires in our guide to AI content personalisation.
Where Distribb fits
We built Distribb for the search half of this problem: finding the queries worth writing for, checking what already ranks, drafting against that structure, and publishing to your CMS. It also handles the internal linking step, which is the one that gets skipped most.
The honest limitation is that Distribb is built for search-driven content. For a homepage, an about page or an ad landing page, it is the wrong tool, and the specs in the table above are better applied by hand. It writes pages meant to be found, not pages meant to be arrived at.
If you write a lot of product-adjacent pages, our AI content writer is the part of the platform that handles drafting against a live SERP rather than from a blank page.
Frequently asked questions
How long should website content be? It depends on the page type. Use the table above: 300 to 600 words for a homepage, 600 to 1,200 for a service page, 1,500 to 3,500 for a guide. Length should follow coverage rather than a target.
How is writing for the web different from other writing? Readers scan rather than read, they arrive mid-way through rather than at the start, and they leave instantly if the first screen does not deliver. That means the answer goes first and the structure has to work when skimmed.
Should I write content myself or hire a writer? Write it yourself for anything requiring knowledge of your product or customers, which is most of your money pages. Hire out the volume work where the expertise is in the writing rather than in the subject.
How many keywords should a page target? One primary phrase, plus the closely related variations that mean the same thing. If two phrases need genuinely different content to answer them, they need two pages.
How often should I update website content? Check pages that rank between positions four and fifteen quarterly, since those move most easily. Leave pages in the top three alone unless something factual has changed.
What to do next
Pick the page you are about to write, find its row in the spec table, and write the headings before the sentences. Check that a stranger reading only your headings would get the point.
If the work is search content at any volume, the research and structure steps are the ones worth automating, and Distribb handles both against live search results. The judgement about what the page is for stays yours.