What content marketing is actually for
Content marketing is answering a question your buyer already has, publishing the answer somewhere you own, and letting it keep working. That last clause is the whole proposition. You are not renting attention for the length of a campaign; you are building an asset a stranger finds at the moment they need it, possibly years after you paid for it.
The honest comparison is with social media, because most Singapore SMEs are choosing between the two. Social is perishable by design: a Reel that performed in March does nothing for you in August, and the moment you stop posting the channel stops producing. Content compounds in the opposite direction. A page that properly answers what something costs in Singapore can earn enquiries every month for years, and it usually gets stronger with age as other pages link to it.
Neither replaces the other. Social creates demand among people who were not looking; content captures demand from people who already are. Most businesses need both and get sold one, because agencies pitch what they are staffed for: a shop full of editors tells you to shoot more video, a shop full of writers tells you to publish more articles. We run both, which is why we are comfortable telling you which one your budget should start with. If nobody knows you exist yet, read the social media side first.
The test that decides which one you need
Check whether people are searching for what you sell. If your category has real monthly volume in Singapore (dental, renovation, corporate services, recruitment, most B2B software), the intent already exists and content is often the cheaper channel because you are catching demand instead of manufacturing it. If nobody is searching yet, which is normal for new categories, content has nothing to catch and the money belongs in social and paid until demand exists. That is a twenty-minute check with keyword data, and it should happen before anyone writes a proposal.
Content marketing is not the same thing as a blog
The blog is usually the least commercially useful part of a content programme and it absorbs most of the budget. Your pricing page, your service pages, a comparison page against the option your buyers are also weighing, a real buyer guide, and the email sequence after a download are all content, and they typically convert several times better than an article about industry trends. We build the commercially loaded pages first and let the blog support them.
What content marketing costs in Singapore
Almost nobody publishes rates here, which suits the agencies. The bands below reflect what brands in this market are typically quoted. They are a planning guide, not our rate card: two articles of identical length can differ in cost by a factor of five depending on how they are researched.
| Engagement | Typical Singapore range | What that usually buys |
|---|---|---|
| Article, commodity | S$150 – S$400 | Eight hundred to twelve hundred words from a generalist working off a brief and a web search. Useful for filling gaps. Rarely enough to outrank an established page or to be worth quoting. |
| Article, specialist | S$500 – S$1,200 | A writer who knows the category, or who interviews someone who does. Original structure, real examples, proper on-page work. Roughly where content starts earning its cost back. |
| Pillar or research piece | S$1,500 – S$3,500 | Long-form built on interviews or original data, with diagrams or comparison tables, a full edit and a distribution plan. Priced per piece because they take weeks. |
| Retainer, light | S$1,500 – S$3,000 / month | A topic plan, two to four pieces a month, on-page optimisation, published into your CMS. Distribution is usually your side at this level. |
| Retainer, full | S$3,500 – S$8,000 / month | Strategy and topic map, four to eight pieces, design, internal linking maintained, email and social repurposing, reporting tied to enquiries rather than sessions. |
| Strategy plus production programme | S$10,000+ per quarter | Category research, an editorial system your team can run, original research or an annual report, multi-format production and localisation. For brands where organic search is a primary channel. |
What actually moves the number
- Research depth. A piece synthesised from page one of Google costs hours. One that reads the regulations, pulls real pricing or checks a claim at source costs days.
- Subject-matter interviews. Every recorded hour of your specialist adds coordination cost, and it is the highest-return line in a content budget.
- Original data. A survey, an anonymised cut of your own numbers, a market teardown. Slow and expensive, and the most reliable way to earn links and citations, because nobody can regenerate it.
- Design. Diagrams, charts and comparison tables raise the price and are often the reason a piece gets referenced elsewhere.
- Localisation. Rewriting for a Mandarin-reading audience is a second writing job, not a conversion step, and should be quoted as one.
- Review burden. Medical, financial and legal work needs technical and compliance review, which adds cycles to every piece and belongs in the estimate from day one.
The cheapest article is usually the most expensive one
The formats, and when each earns its keep
Format follows the job. Most disappointing content programmes are one format, usually blog articles, applied to every objective including several it is bad at.
| Format | What it is genuinely good at | When it is not worth it |
|---|---|---|
| Pillar guide | Owning a broad category term and giving smaller pages something to link to. The anchor of a topic cluster, and the piece most likely to be cited by other sites. | Before you have anything to link into it. A pillar with no supporting pages is just a long article. |
| Comparison and alternatives page | Catching buyers at the end of the decision, weighing you against a named option. Usually the highest-converting page in a content programme. | When you cannot be fair. A comparison that only flatters you is obvious to readers, and is exactly what assistants decline to quote. |
| Buyer guide and pricing page | Answering what everyone in your category refuses to answer publicly: costs, timelines, what changes the number. Attracts serious buyers and filters out the rest. | Almost never. If pricing is genuinely bespoke, publish the ranges and the variables rather than nothing. |
| Case study | Proof. The format sales teams actually use, and the one that closes a buyer who needs to see a business like theirs. | When you cannot name the client, share a number, or describe the real problem. A vague case study is worse than none. |
| Email sequence | Converting attention you already earned. The cheapest content per enquiry you will produce, because the audience opted in. | When nothing feeds the list. A sequence with no acquisition in front of it is an unread document. |
| Landing page | Turning one intent into an enquiry: one audience, one offer, one action, matched to the ad or article that sent them. | As a substitute for a service page. Landing pages rarely rank; they convert traffic you brought. |
| Whitepaper or original report | Authority, links and press in categories where nobody has published real numbers. Reusable across sales, email and social for a year. | When the data is thin. A PDF restating public knowledge behind a form annoys the exact people you wanted. |
Writing for AI assistants as well as Google
A growing share of buyer research now happens inside an assistant: ChatGPT, Claude, Perplexity, Gemini, or Google’s AI overviews sitting above the normal results. That changes what a page has to do, and a great deal of confident nonsense is being sold about it. Here is what is mechanical, and where the honest answer is that nobody knows yet.
How an assistant gets your page in the first place
When an assistant answers a question about the real world, it is usually not recalling your site from training. It runs a search, retrieves a few pages, fetches them, and writes from what it read. So conventional search visibility is the entry ticket: a page no index holds and nothing links to cannot be retrieved, and therefore cannot be cited, however well it is written. Google’s AI overviews draw on the ordinary Google index, so there is no separate door there either. SEO fundamentals were not replaced. They became the qualifying round.
Structure and citable specifics beat keyword density
Retrieval works on passages, not whole documents. A section gets pulled out on its own, without the three paragraphs above it, and either answers the heading in front of it or does not. In practice: phrase headings as the question a person would ask, answer it in the first two sentences beneath, and make every section stand up when read cold. Specifics are the currency. A price range, a statutory deadline, a named tool, a number from your own operations: those can be attributed, so they get quoted. “A leading provider of innovative solutions” can be attributed to nobody, which is why nothing ever quotes it.
JavaScript-rendered content is invisible to most AI crawlers
This one silently costs businesses the most. Googlebot renders JavaScript, so an app that assembles its content in the browser can still rank. Most AI crawlers do not: they read the raw HTML the server returns. If your text only appears after a client-side fetch, they see an empty shell. Server-side rendering or prerendering fixes it, and it is worth checking before you spend anything on content. View the page source, and if the article body is not in there as text, half your audience is reading blindfolded.
The crawler you may be blocking by accident
AI crawlers are named and controllable: GPTBot and OAI-SearchBot for OpenAI, ClaudeBot and Claude-SearchBot for Anthropic, PerplexityBot for Perplexity, Google-Extended as a control for Gemini. Two things get confused. First, several CDNs and hosts now block AI crawlers by default, so a site can be invisible to assistants without anyone deciding that, which makes robots.txt and your CDN bot rules worth an actual look. Second, Google-Extended governs Gemini and related grounding, not your presence in AI overviews, which come from ordinary Google indexing: there is no setting that keeps you in Google search and out of the AI answer above it. Blocking is a legitimate choice. Make it deliberately, knowing that opting out of being trained on is also opting out of being cited.
What llms.txt is, and how seriously to take it
An llms.txt file is a proposal from late 2024: a markdown file at the root of your domain listing your most important pages with a line of description each, sometimes with a fuller llms-full.txt carrying the content itself. A sitemap written for language models rather than crawlers. It is sensible, it takes about an hour, and we add one because it is cheap and harmless. What we will not tell you is that it works: no major assistant has publicly confirmed it reads the file, and anyone selling a generative engine optimisation package whose centrepiece is llms.txt is selling you a text file.
What nobody can honestly promise yet
How we work
We interview you before we write anything
The difference between a page anyone could have written and a page only your business could publish is almost always already in someone’s head. So we record a proper conversation, usually forty-five to sixty minutes per topic cluster, with whoever does the work: the founder, the senior technician, the person who answers the phone. Out of that comes the raw material, the objection you hear weekly, the number you always end up quoting, the mistake customers keep making. None of it is available to a writer researching online, which is exactly why it is worth publishing.
Drafting, then a real edit
We use AI in production and are not coy about it: outlines, variants, first drafts, table structures. Then a person edits every piece, checks every number against a source, and deletes anything we cannot verify. That pass is why we ship fewer pieces per month than a content mill and cost more per piece. It is the honest trade: unedited AI output is not badly written, it is generic, and generic is the one thing that no longer earns anything from Google or an assistant.
A cadence, and a refresh half
Four solid pieces a month without gaps beats twenty in a quarter followed by silence, and buyers do notice a blog whose last post is fourteen months old. About a third of the calendar goes to updating pages that already exist. That is usually the cheapest ranking available to you, and it is the line most agencies skip because new pieces look better on a report.
What we need from you
Roughly two hours a month: one interview, one review round. Plus CMS access and one named person who can approve a draft without a committee. Approval chains, not ideas, are the usual reason a content programme quietly stops publishing.
How to choose a content agency
Most buying advice in this category is written by agencies and describes themselves. Here is the version we would give a friend, including the parts that do not favour us.
- Ask who writes it, by name, and read their work. Content is one of the few services where the individual matters more than the agency.
- Ask how they will get the subject knowledge. If the answer is that they research online, your pages come from the same sources as your competitors’ pages and will read that way. Interviews, or access to a real specialist, are the only two good answers.
- Ask what happens after publishing. Internal linking, refreshes, repurposing into email and social, and a review of what actually got traffic. A programme that only produces is half a programme.
- Ask to see the topic plan and the intent behind it. Every topic should map to a term people search or a question sales genuinely hears. A plan that cannot show you that was written from imagination.
- Be suspicious of guarantees, old and new. Guaranteed rankings were always a bad sign. Guaranteed AI citations are the 2026 version of the same sentence.
When not to hire a content agency at all
Where we are genuinely a poor fit
If you need newsroom volume, thirty or more pieces a month across several markets, you need a managing editor and a bench of writers, and we are the wrong size. Same answer if you need an in-house compliance desk for regulated financial or medical publishing. We are a studio: a smaller number of pages that are worth reading, and we are part of Digital 9 Labs, so when the content needs a site, a booking flow or tracking that actually attributes an enquiry, the same group builds it. The rest of what we do is on the studio home page.