SEO Style Guide: What Writers Need Written Down

An SEO style guide is not a brand style guide with a keyword bolted on, and it isn't a content strategy document either. A brand guide governs tone and logo use; a strategy document decides which topics to cover and when. A style guide governs the mechanics every writer, and every AI-assisted draft, has to get right before a page goes live: how titles and headings are built, how internal links get placed, which terms are used consistently, how alt text gets written, when a claim needs a source, and how authorship gets disclosed. Skip it, and quality becomes a function of which writer happened to draft that week's page.

This matters more, not less, once part of the drafting is AI-assisted. A model will invent a heading structure, skip alt text, or state a fabricated statistic with total confidence, because nothing told it not to. The style guide is the document that tells it not to, and tells the human editor exactly what to check before publishing.

What an SEO Style Guide Actually Covers

A style guide answers a narrower question than a strategy document does: given a topic that has already been chosen, how should this specific page be written and marked up? It has no opinion on which keywords to target or which cluster a post belongs to — those decisions happen earlier, in planning. Folding them into the same document is why so many 'style guides' turn into decks nobody reopens after the kickoff meeting.

What belongs in scope is anything a writer has to decide the same way every time, regardless of topic:

• How titles and headings are capitalized and structured
• How and when to link, internally and to outside sources
• Which terms, spellings, and names are used consistently
• How images get alt text and file names
• When a claim needs a citation
• How authorship and AI assistance are disclosed

Title and Heading Rules

Every page gets exactly one H1, close enough to the title tag that a reader who clicks from search results doesn't feel bait-and-switched. Sixty characters is a reasonable working limit for the title itself — not an enforced rule, but search results truncate what doesn't fit the display width, and a title cut off mid-word looks unfinished. The target phrase should sit naturally near the front, where a scanning eye lands first, not because front-loading is a ranking hack, but because that's simply where attention goes.

Headings after the H1 need to nest in order — an H3 sits under an H2, never skipping straight to H4 because it looked right visually. That hierarchy is read as a structural outline by screen readers and crawlers, not just as font sizes, so picking a heading level for its size breaks the outline for both audiences. Headings should describe what a section contains, not tease it: 'Pricing Tiers Compared' tells a reader something; 'Let's Talk Numbers' doesn't.

A brand publishing in more than one language needs the capitalization convention stated per language, because English's Title Case habit doesn't transfer. English headings can reasonably use Title Case; Spanish, Italian, Portuguese, and French headings should use sentence case — first word and proper nouns only. Writers who default to English convention across every language are the most common source of inconsistency, and it's rarely caught, because non-English pages get read less carefully.

• One H1 per page, matching the title tag's intent
• Sequential heading levels, no skipped tiers
• Headings a reader could use to skim the page alone
• A stated capitalization rule for every publishing language

Linking Rules: Internal and External

Internal links are where a style guide earns its keep, because they're the one on-page element a writer working alone genuinely can't keep consistent without a written rule. Decide in advance: a rough range for how many internal links suits a post this length, whether the same anchor text pointing at the same page repeatedly across the site is acceptable (usually not — it reads as a manufactured pattern rather than organic linking), and whether every new post must link back to at least one page already in its cluster, so nothing publishes as an orphan.

External links need the opposite instinct: writers under-link out of a fear that sending a reader elsewhere loses the visit, when a citation to a primary source — the actual documentation, the actual tool, the actual data — is what makes a claim checkable and is a quality signal in itself. The guide should say when an outbound link is expected, such as anywhere a specific claim or third-party tool is named, and what attribute it gets: a plain link for an editorial citation, a sponsored attribute for a paid placement, nofollow for untrusted or user-generated content. Someone should periodically check that old outbound links still resolve, because a citation now leading to a dead page reads as neglect.

• A stated range for internal links per post, not a hard minimum
• No repeated exact-match anchor text into the same target page
• A rule for which link attribute applies when
• An owner for checking that citations haven't rotted

Terminology and Consistency

Every brand accumulates terms that get written three different ways unless someone settles the one correct form: 'e-commerce,' 'ecommerce,' or something else entirely. A short glossary, often no more than thirty entries, settles this once instead of leaving it to a copyeditor to catch, inconsistently, after the fact. The same list should note which acronyms need spelling out on first use in a given piece, and which are common enough to leave unexpanded.

Terminology consistency also covers the phrases a guide actively bans, not just the ones it standardizes: stock filler like 'in today's fast-paced digital landscape,' absolute claims like 'guaranteed to rank,' and hedge words used purely as padding. None of this is about house voice in the branding sense — it's about removing the small inconsistencies that make a hundred-post archive read as though it never had an editor in common.

• A glossary of contested spellings and capitalizations, kept short
• A rule for which acronyms need first-use expansion
• A list of banned filler phrases and absolute claims

Alt Text and Image Conventions

Alt text exists first for accessibility — it's what a screen reader announces in place of an image — and its usefulness to image search is a byproduct of writing it well for that purpose, not a separate skill. A good description names what's actually in the frame, concretely enough that someone who can't see it loses nothing: 'bar chart comparing click-through rate by result position' says something real; 'SEO image,' or the keyword repeated verbatim, says nothing and reads as stuffing to anyone auditing the page. Purely decorative images — a divider, a background texture — get an empty alt attribute, not a description forced onto something that isn't conveying information.

The guide should also fix a file-naming convention (hyphenated and descriptive, not a camera-generated string) and a rule for when an image needs a visible caption versus when alt text alone is enough. A caption is for information a reader benefits from seeing even if the image never loads — a data source, a date, a unit of measurement — while alt text is written for the case where the image doesn't render at all.

• Alt text describes the image concretely, not a second place for the keyword
• Decorative images get an empty alt attribute
• File names are descriptive and hyphenated

Sourcing and Fact-Checking Rules

The rule that matters most here is simple to state and hard to enforce without writing it down: any number in a draft — a percentage, a claimed ranking-factor weight, a promised timeframe for results — needs a named, checkable source, or it comes out entirely. This is the rule most likely to bend under deadline pressure, and exactly the one an AI-assisted draft will violate without anyone noticing, because a plausible-sounding statistic is easy for a model to produce and hard to tell apart from a genuinely sourced one just by reading it. 'Studies show that…' with no study named isn't a citation; it's a tell the number was invented to sound authoritative.

Separately, the guide should distinguish mechanics a writer can state as settled fact — how a redirect passes signal, what a canonical tag asks a crawler to do — from claims about how a specific platform currently behaves, which change over time and need a date or a link to that platform's own documentation rather than a flat assertion. A sentence that attributes intent or policy to a named platform should be checked against that platform's own published guidance before it goes out, not assumed from what used to be true or from what a competitor's blog once claimed.

• Any specific number needs a named source or gets removed
• 'Studies show' with no study named is a red flag, not a citation
• Claims about current platform behavior are checked against that platform's own documentation

Bylines, Authorship, and AI Disclosure

A byline is a trust signal before it's anything else, and it only works if it's consistent and real. The guide should fix the format — name, one relevant credential, a link to a bio page — and apply it the same way whether a post was drafted by a staff writer, a freelancer, or a model working alongside a human editor. Publicly documented rater guidelines ask evaluators to look for evidence of who wrote a page and why they're positioned to write it, which is reason enough to get this right beyond simple honesty.

Where AI assistance was used in drafting, the guide needs a clear policy on whether and how that gets disclosed, applied the same way on every post instead of decided case by case under deadline. The safer default is to disclose the workflow — drafted with AI assistance, edited and fact-checked by a named human — rather than to either overclaim full human authorship or bury a disclosure where nobody reads it. Writing the policy down once means no individual editor has to decide it post by post, which is exactly when policies drift.

• A fixed byline format applied identically regardless of who drafted the post
• A named human accountable for fact-checking, even on AI-assisted drafts
• One disclosure policy applied consistently, not decided case by case

Keeping the Guide Alive: Versioning and Enforcement

A style guide that never changes is either perfect or ignored, and it is never the first one. Search engines update their guidance, a brand adds a product line that needs a new glossary entry, an editor spots a recurring mistake that needs a rule instead of a repeated correction — all of it means the document needs a version number and a short changelog, not a single 'final' file in a shared drive. One person should own it, not because no one else can propose a change, but because an unowned guide accumulates contradictory edits from whoever touched it last.

Enforcement works better as a short checklist attached to the publishing workflow than as a document writers are expected to hold in memory in full. A five-line pre-publish checklist — heading hierarchy correct, internal link present, alt text written, numbers sourced or removed, byline and disclosure applied — catches most of what the longer guide covers, at the exact moment it matters, without asking anyone to hold forty pages in their head while drafting.

• A version number and changelog instead of a single static file
• One named owner for changes
• A short pre-publish checklist that operationalizes the rules