Editorial Guidelines

Who can write

The SEM Dispatch is open to contributions, but open does not mean anyone. We publish working practitioners with genuine depth in search marketing. If you are still learning the channel, this is a publication to read, not yet to write for.

To be accepted as a contributor you need three things.

• At least 1 year hands-on in the channel. Time spent running real accounts, with real budgets and real accountability for the outcome. Years adjacent to the work, in sales or general marketing, do not count towards the bar.
• Genuine, original work to share. A method, an argument, a post-mortem or an analysis that comes from your own accounts, not recycled takes or AI filler. You need no personal brand, published clips or conference reputation. The bar is the quality of your work, never who you know, and we read and approve every author personally before they join.
• A commitment to file at least once a month. One filing a month is the minimum for an active byline. It is light enough to sit beside a full-time job, but it is a genuine commitment: this is a working desk, not a place to publish once and disappear.

Who this is, and is not, for. Bylines are for practitioners who run search marketing: agencies managing client accounts, and in-house teams running their own. They are not a marketing channel for the companies that sell to those marketers. If you are a martech or ad-tech vendor, a platform, or a service for marketers, the right route is a commercial partnership, not a byline: write to info@byltmedia.com.

What we publish

One test sits above every other: a competent practitioner should finish the piece able to act on it, learn something real, or implement it on Monday morning. If a reader cannot do anything differently afterwards, it is not ready. We publish four kinds of article, and each carries its own bar.

• News. Not a summary of what happened. The announcement is where the piece starts, not where it ends. Tell the reader what it means for them, why it matters, and how it changes the way they should work. If they could get the same from the original announcement, we do not need it.
• Analysis. Data-led, drawn from your own accounts wherever possible, with your working shown: the sample, the time period and the assumptions. First-party, proprietary data is the strongest thing you can bring to the masthead, so bring it.
• Guides. Actionable strategy a reader can put to work: the concrete steps, the reasoning behind them, and real examples and screenshots from accounts you have run. Method and evidence, never theory in the abstract.
• Opinion and thought leadership.A clearly qualified point of view, argued from evidence rather than asserted. Welcome and wanted, as long as it moves the reader’s thinking rather than restating what the channel already believes.

On length, we set no word count. A piece is as long as the idea genuinely needs and not a sentence longer. Say the valuable thing in full, then stop. A short, dense filing that earns every line is worth more than a long one padded to look thorough, and filler is the one length mistake we will always send back.

Fresh angles, not repeats

We cover a subject once, not three times in a month. If we have just run a strong piece on something, a second filing that lands in the same place adds nothing for the reader, and we will hold it back or ask you to find a new angle.

Disagreement is the welcome exception. If we published a piece arguing that AI Max works and you have run the accounts and found it does not, that is not a repeat: it is the other half of the debate, and we want it. A genuinely different position, a contrary reading of the same data, or a fresh angle on a tired subject is original work. What we treat as a repeat is a restatement, the same argument reached the same way, saying what has already been said. Bring the evidence either way: a contrary view earns its place by being argued from what you have seen, not asserted against the grain.

Using AI

The question we are asked most is whether you can use AI to write for us. The honest answer turns on what you use it for. To polish, yes. To think for you, no.

• To polish, freely. Tightening a sentence, checking your grammar, trimming a paragraph that runs long: a language model is a perfectly good editor of work that is already yours, and we have no objection to it. Plenty of strong writers use one this way.
• To generate, never. What will not run is a draft the model wrote and you lightly signed off. We are not short of content a machine can produce on demand, and neither is the reader. A search engine answers those questions already.

What earns a place here is the thing only you have: your opinion, your hard-won expertise, and the first-party data from accounts you have actually run. That is what a reader cannot find anywhere else, what they remember you for, and, quietly, what wins you the clients reading over their shoulder. A support article recycled from a platform’s own help centre does none of that. The reader can read Google’s documentation without us, so a filing that only restates it adds nothing, and the desk will decline it.

What we decline

Some submissions are declined no matter how well written they are, because they do not serve the reader. We do not publish:

• Sales material of any kind. Anything written to promote your product, agency, course or platform, however lightly dressed as editorial. If the piece works as an advert, it is an advert, and it is not for us.
• Beginner explainers that any search engine would answer in seconds.
• Recycled takes with no original insight, data or experience behind them, including support articles restated from a platform’s own help centre.
• Drafts generated by a language model and submitted without substantial human expertise, judgement and editing.
• The same article twice: a filing that duplicates a piece we have run recently, without a genuinely new angle or a contrary view argued from evidence.

The review process

Open does not mean unedited, but our touch is light. Every piece is reviewed by a desk editor who has done the job, we tighten where it helps and add design where a filing needs it, and we never rewrite your voice. The flow is the same for everyone.

1. Apply. You tell us who you are, your experience and the topics you want to cover, and link to work that shows your depth.
2. Pitch. Once you are approved as a contributor, you send a short outline of a specific article from your dashboard. An editor responds within five working days.
3. Draft. When a pitch is accepted, you draft the piece in the dashboard as Markdown, using the house marks (:::callout, :::table, :::faq and so on) for rich blocks, and submit it for review.
4. Editorial review. A desk editor reviews for accuracy, originality and style, fact-checks the claims, and may ask for revisions before it is cleared.
5. Publish. Approved pieces are committed as git-versioned content and published under your byline, with your contributor profile attached.

House style

A consistent house style keeps the publication readable and the focus on your argument. The essentials:

• British English throughout. Optimise, not optimize. Behaviour, not behavior. Colour, not color.
• No em dashes. Use a comma, a colon, or a middot (·) instead. This is a strict house rule, not a preference.
• Plain, confident prose. Short sentences. No jargon for its own sake, and no filler.
• Show the evidence. Claims need numbers, examples or sources standing behind them.
• Title case for headlines, sentence case for subheadings.
• Define acronyms on first use, even the ones you think everyone knows.

House marks

The body is Markdown. For the publication’s richer blocks you do not need to write any code: wrap a block in a :::namefence, write plain Markdown inside, and close it with a line that is exactly :::. The desk compiles it into the house component, and the dashboard Preview shows the result before you submit. There are six marks.

• :::callout for the single point a skim-reader must not miss. The words after the name become its label.
• :::quote for a line worth lifting out; the words after the name are the attribution.
• :::table for figures that read better as a grid. Write an ordinary Markdown table inside, and right-align a numbers column by putting a colon on the right of its divider dashes (---:).
• :::faq for the real questions you are asked. Start each with a bold question line, then write the answer below it. It also produces Google’s FAQ structured data.
• :::sources for the references behind the piece: a Markdown list of [label](https://link).
• :::figure for an image or a captioned placeholder: optional src:, alt:, label: and ratio: lines, then the caption.

Your TL;DR and key takeaways are separate fields in the editor, above the body, so there is no mark for them. Older articles use the longer explicit form (<Faq items=… />); you no longer need to. If a block prints as plain ::: text in Preview, the name is misspelt or the closing ::: is missing.

Images and the safe zone

One picture does the work of four. The image you attach to an article is the same file the front page, the section cards and the search and social previews all draw from, and each crops it to a different shape. Compose for that and your picture always looks deliberate; ignore it and a crop will cut a face or a logo in half.

Supply a landscape 16:9 image, at least 1600px on the long edge. Everything that matters, the subject, faces, logos and any words, belongs in the centre: 15% in from each side and 10% from the top and bottom. The square-ish thumbnail crop is the real danger, taking 12.5% off each side, so the outer edges are no place for anything you cannot afford to lose. The uploader draws this safe zone over your image as you place it, so you can see at a glance whether the subject sits safely inside it.

The strongest hero carries no text at all. The page sets your headline in type beside the image, so words baked into the photo only compete with it and get clipped in the narrow crops. Let the picture be a picture.

Every photo gets the same Picture Desk treatment automatically, grayscale with a fine newsprint screen, so the colour stays on the page and never on the photograph. Generate or shoot clean and let the page do the look. The full template, with the bleed map and the crop demonstrations, is one click away.

Open the full Picture Desk template →

To brief an image generator, this prompt produces a frame that drops straight into the safe zone:

16:9 editorial photograph. Single subject, centred, filling the middle 60-70% of the frame. Generous clean negative space on all four edges (safe bleed). Strong tonal contrast, real blacks and whites (it prints black and white). No text, no captions, no logos, nothing important near the edges.

Evidence and sourcing

Results are only worth printing if a reader can believe them. Every data claim must be backed by a screenshot from the platform it came from, the dashboard, the report or the interface. A number with no screenshot behind it does not run. Anonymise account data by default, and be clear about the sample size and the time period it covers. You may name a client or their results only with their written permission; without it, the data stays anonymous. Show the evidence and the reader will trust the argument.

Everything you file must be your own work, written for The SEM Dispatch and not published elsewhere first. Link to primary sources wherever possible, and if you cite a statistic, link to where it came from. We fact-check claims during review, and we will ask you to substantiate anything that looks shaky before it goes to press.

Disclosure and conflicts

Trust is the only currency a community publication has, so we are strict about transparency. Declare your company and any commercial relationship relevant to what you are writing about. If you work for, advise, or hold equity in a tool you mention, say so in the piece.

Where a contributor writes alongside their firm, we treat it openly. A piece may carry a clearly labelled bureau or sponsored framing, with the company named and its interest stated, so the reader always knows who is speaking and why. What we never do is let a commercial interest pass as neutral editorial. Undisclosed conflicts are grounds for removing a piece and a contributor.

How to submit

Start by applying to write. Approved contributors get a dashboard where they pitch, draft and submit for review, and where the editorial conversation happens. You do not need to write the full article to apply: a clear sense of your experience and two or three ideas is enough for an editor to judge the fit.