Five quotes for the same job, one 1,500-word post on a topic your sales team keeps getting asked about: $75, $180, $400, $950, $1,400. Every one says “SEO-optimised, well-researched, original”. The cheapest will start tomorrow. The most expensive sent back three questions before quoting, one of which you could not answer.

The spread means something, but not what the price sheets imply. Per-word pricing is the wrong unit for outsourcing SEO content, because word count is the one input with almost no relationship to what the work costs. What you are actually buying is hours of research and someone’s willingness to be accountable for the claims, and both scale with the difficulty of the topic, not the length of the piece. A competent writer produces 1,500 words in three hours when the topic is already fully documented on page one of Google. The same writer needs fifteen hours when the piece means reading a 40-page regulation, interviewing two of your engineers, and being right about something a regulator may ask about later.

The procurement answer, before the detail. In 2026, budget roughly $400 to $1,200 for a 1,500-word piece on a topic where the writer must understand your product or your industry, and $1,500 upward where it needs interviews, primary documents or original analysis. Below about $250 you are buying assembly, not research. The rest of this is how to spend inside those bands without becoming the buyer who pays $1,200 and still gets a rewrite of the first three results.

Why the same 1,500 words costs $150 or $1,500

Four things drive the cost. None is length.

1. Subject-matter expertise. Not “did research” but “already knows, or can be trusted to learn without producing plausible nonsense”. A writer who has run a revenue cycle team covers denial codes in four hours. A generalist takes twelve and still gets the appeal timelines wrong, because the wrong version is what most of the internet says. You pay for the expertise, or for the hours spent imitating it.

2. Original inputs. Interviews, a dataset, a document nobody has summarised, a product the writer actually installed. An interview is not the 30 minutes on the call: it is scheduling, preparing questions worth a customer’s time, the transcript, the follow-up when a quote needs clarifying, and approval from whoever said it. Two interviews add six to eight hours, the largest single swing factor in any quote you will receive.

3. Liability and review burden. In health, finance, legal, insurance, employment and safety topics someone has to own the claim: a source list, a clean chain from statement to citation, usually a named reviewer. Suppliers in these areas price the review in because they carry it. The ones who do not cannot tell you where a number came from three months later, exactly when someone asks.

4. Revision cycles. The invisible cost, and the one you control most. A piece that goes four rounds because the brief was vague costs the supplier double and costs you more than that in internal review time. Good suppliers price defensively for it, which is part of that $1,400 quote, or cap revisions in the contract. Bad ones absorb it once, then quietly cut the research on everything after.

Three of those four are settled before a word gets written. That is why per-word pricing keeps disappointing buyers: it prices the one stage that is now nearly free.

What outsourced content actually costs, by the published rate data

The best-documented public source is the Editorial Freelancers Association’s 2026 Editorial Rate Chart: self-reported, over 1,100 respondents representing more than a third of EFA membership, collected November 2025 to mid-January 2026 and covering 2025 rates. Membership skews toward book, academic and publishing work, so read it as a well-documented slice rather than the whole market, and still far more transparent than anything else you will be quoted.

Its writing lines, converted into the price of one 1,500-word piece:

• Ghostwriting, articles and essays: 15 to 30 cents a word, $75 to $100 an hour. $225 to $450.
• Work-for-hire, articles and essays: 25 to 45 cents a word, $70 to $100 an hour. $375 to $675.
• Ghostwriting, technical: 87.5 cents to $1.08 a word, $75 to $126 an hour. $1,313 to $1,620.
• Ghostwriting, medical: $1.00 to $1.50 a word, $100 to $126 an hour. $1,500 to $2,250.
• Fact-checking $50 to $60 an hour, research $47.50 to $60. Both hourly, because nobody has ever successfully priced verification by the word.

Read the hourly columns next to the per-word columns and the argument makes itself. The same survey shows a tenfold per-word spread and roughly a 1.5x hourly spread. Hourly rates barely move. What moves is how many hours the topic demands. A $450 article and a $2,250 article are not different grades of writing so much as five hours of work versus eighteen.

Buyer-side figures are murkier. The most-cited, WebFX’s pricing page, reports $150 to $600 per blog post and $500 to $2,500 for long-form articles from a survey of 350 digital marketers. Handle with care: an agency that sells content published it, no fielding date is given, and respondents are not identified as buyers, agencies or freelancers. If a vendor quotes a market average and cannot name the sample and the fielding date, the number is decoration.

Buy tiers, not words

Replace the per-word line item with three tiers, and pick the tier when you pick the topic, before anyone quotes.

Tier 1, commodity. $150 to $400. Definitional pages where the correct answer is already fully public and your job is to say it clearly: glossary entries, “what is X”, how-to on a widely documented tool. Three to five hours. What it buys is a competent restatement of the consensus, which ranks only where competition is weak. Value on its own: near zero, and in the wrong conditions negative.

Tier 2, expertise. $500 to $1,200. The writer must understand your product, your buyer or a regulated domain. They read your docs, get 30 minutes with an expert on your side, use primary sources, and explain a tradeoff rather than list features. Eight to twelve hours. Most B2B content belongs here, and here is where the disappointment happens, because buyers pay tier 2 prices with tier 1 briefs and tier 1 access. Worth: pages your sales team sends to prospects unprompted, which is a better test than rankings.

Tier 3, original research. $1,500 to $5,000 and up. Two to five named interviews, an original survey or dataset, a teardown of something real, analysis that did not exist before you paid for it. Fifteen to forty hours. This is the tier that earns citations, links and mentions in AI-generated answers, because it is the only one holding information not already in circulation. If your average deal is $40,000, one such piece that influences a single deal a year has paid for itself six times over. Run four to six a year, not forty.

A sensible mix at volume is roughly 60% tier 2, 25% tier 1, 15% tier 3. If yours is 90% tier 1, you have a publishing schedule, not a content programme.

Why the bottom tier’s expected value is negative, not zero

The comforting assumption about cheap content is that a bad piece wastes money and nothing more. Not so. Google’s spam policies define scaled content abuse as “when many pages are generated for the primary purpose of manipulating search rankings and not helping users”, giving as an example “using generative AI tools or other similar tools to generate many pages without adding value for users”. Google says it detects violations “both through automated systems and, as needed, human review that can result in a manual action”, and that offending sites “may rank lower in results or not appear in results at all”.

Read that precisely, because it gets misquoted in both directions. AI use is not the violation. Scale without added value is, whatever produced the pages. Forty thin pages a month is the named abuse, and enforcement lands at site level, not page level. So the downside of the cheap tier is not the $150. It is that a few hundred thin pages can drag down the pages you paid properly for.

Google’s guidance for creators also asks whether content “provides original information, reporting, research, or analysis” and whether it “is mass-produced by or outsourced to a large number of creators”. Those two questions describe exactly the supply chain a per-word buyer builds by default.

The brief is where the money leaks

In my experience the biggest cause of expensive content is not a bad writer. It is a brief containing a keyword, a word count and a due date, so the writer guessed at the angle, guessed wrong, and two people now spend four hours each fixing it. That rework never shows in the content budget, because it arrives as internal salary rather than an invoice.

A brief that prevents it takes 25 minutes and covers twelve things:

• The one decision this page helps the reader make, in a sentence. If you cannot write it, do not commission the piece yet.
• Who the reader is: job title, what they already know, what would insult them, plus the primary query and three to five long-tail phrasings people actually type.
• What the top results say and where they are thin. This is the gap you are paying to fill, and naming it is your job, not the writer’s.
• Non-negotiable facts and source documents, linked. Internal docs and support tickets count double.
• Who can be interviewed: name, email, booking link. Access, not permission in principle.
• Claims we cannot make, from legal or compliance. Cheaper up front than in round three.
• Internal links to include, the one page this piece exists to send traffic to, and format constraints: a word range rather than a target, headings, whether tables or screenshots are needed.
• Acceptance criteria: three to five bullets describing a pass, written before the draft arrives.
• Revision policy and deadline, with the named reviewer’s turnaround committed in writing. Late client review is the commonest cause of late delivery, and suppliers are too polite to say so.

One rule to adopt with it: word count is a constraint, never a goal. The moment a writer is paid to reach 1,500 words, you have paid for 300 words you do not want.

A paid test that actually tests something

Free samples tell you nothing, and portfolio pieces tell you what an editor made of someone’s draft two years ago. Run a paid test instead, designed so it can fail. Pick a topic you know cold, ideally one where the common online answer is subtly wrong. Pay the full tier rate for one piece, give them your real brief and the access a regular writer would get. Judge four things, in this order:

• Did they catch the thing a non-expert would miss? That is the whole test. If the piece repeats the wrong internet consensus confidently, no amount of clean prose saves it.
• Did they ask a question before starting? Writers who research ask something. Writers who assemble do not need to.
• Is any source primary? A statute, a filing, a vendor’s own documentation, a dataset, a person. If every citation is another blog post, no research happened, only reading.
• Did they push back on the brief? Best single signal available. A supplier who tells you your angle is wrong, and is right, is worth two who deliver exactly what you asked for.

Run three candidates in parallel on the same brief. At tier 2 that is $1,500 to $3,600 of testing, which sounds expensive until you price the alternative: finding out after fifteen pieces and four months that your supplier cannot do the work.

AI is in your supply chain now. Write the contract accordingly

Assume every draft has had some machine involvement. CMI and MarketingProfs research fielded 25 June to 16 August 2024 across 980 B2B marketers put generative AI use at 81% of teams, and the direction since has been one way. A blanket “no AI” clause is unenforceable theatre that mostly punishes honest suppliers. Contract for what you actually care about:

• Disclosure by stage, not in general. Require a statement of where generative tools are used: research, outlining, drafting, editing. “We use it for outlining and never for factual claims” is workable. “We don’t use AI”, from a volume supplier, is usually just a less accurate answer.
• A source list with every piece, every factual claim mapped to a link. Cheap to require, and it quietly makes generated-and-unchecked work impossible to hand over.
• A named human author who stays reachable and will answer a question about the piece twelve months from now. This clause does the most work, because it forces someone to have actually understood the topic.
• A factual accuracy warranty, named review sign-off on regulated topics, and a throughput cap per writer. Someone delivering fifteen researched pieces a week is not researching.
• A reader-facing disclosure policy. Google’s guidance on creating helpful content asks whether “the use of automation, including AI-generation, self-evident to visitors through disclosures”. Decide yours deliberately.

As for spotting a piece that was generated rather than researched: do not buy a detector, audit the piece. Ten minutes, and it produces evidence you can act on.

Open every citation and check the page exists and says what the draft claims; invented or mismatched citations are the most reliable tell there is. Ask whether any source is primary: a filing, a specification, a regulation, a dataset, a transcript, your internal documents. Then look for what could only be known by someone who did the work. A screenshot of the step where the setup wizard fails. A number from your dashboard. A sentence a customer said. The reason the obvious approach stops working at 200 users. Generated text is strong on structure and consensus, weak on friction, and friction is only met by people who did the thing. A draft with none in it is a summary of the internet, and you had that for free.

In-house, agency, freelancer, offshore: the real comparison at volume

At twenty researched pieces a month the four models separate clearly. Arithmetic first: Bureau of Labor Statistics data puts median pay for writers and authors at $76,910 as of May 2025, and employer benefit costs in private industry at 30.0% of total compensation as of June 2026, so a writer on median salary costs roughly $110,000 loaded, before tools, recruiting or management. A good in-house writer doing genuinely researched work delivers six to eight pieces a month, not twenty. Call it $1,150 to $1,530 a piece before anyone manages them.

Model	Typical all-in cost per researched 1,500-word piece	What you are really buying	Where it breaks
In-house writer	$1,150 to $1,530 plus management	Product knowledge, fast access to colleagues, institutional memory	Capacity ceiling, single point of failure, narrow topic range
Agency	$800 to $2,500	Capacity, project management, one accountable contact, cover when someone leaves	The writer is often a freelancer on a fraction of the fee; quality tracks the account manager, not the logo
Direct freelancer roster	$450 to $1,200	Best cost per unit of quality at tier 2, and a direct line to the person doing the work	You supply the management. Budget three to five hours a week of an internal editor or it degrades quietly
Offshore or nearshore team	$150 to $500	Volume, research support, page updates, formatting, localisation, consistent process	Struggles where the piece requires interviewing your customers or owning a regulated claim, unless that capability is built deliberately

One correction on that last row, because buyers get it wrong routinely: the assumption that offshore means tier 1 only is lazy. An offshore writer with real domain training in, say, medical billing beats a domestic generalist on that topic every time, and process consistency across 400 existing pages is often worth more than the per-piece saving. What travels badly is not skill. It is access to your customers and your regulators.

The structure I would recommend for most companies buying at volume: one in-house editor who owns briefs, acceptance and the source-of-truth documents, three to five specialist freelancers for tiers 2 and 3, and an offshore or managed team for tier 1 volume, updates and research support. The editor is the load-bearing hire. Without one, every model in that table produces the same disappointment, and you will conclude the problem was the price.

Two questions buyers ask next

What is a fair revision policy? Two rounds, where a round means consolidated feedback delivered at once. Anything changing the angle, audience or core argument after brief approval is new scope and should be priced. Most revision disputes are brief disputes arriving late.

How do I set the budget at all? Work backwards from the value of a customer: average deal value times close rate on organic leads. If a tier 2 piece plausibly influences two deals a year and a deal is worth $8,000, a $900 piece is not an expense to minimise. Companies that cannot compute that number are the ones arguing about cents per word. And pay on time: net 60 on a $600 invoice is how you lose good freelancers.

If you have the budget but not the editorial capacity to run it, that is a staffing problem rather than a writing one. AB7 Solutions places dedicated remote content and SEO professionals, editors, researchers and subject-matter writers who work as an extension of your team on your briefs and your review process, alongside our SEO, AEO and GEO work for companies trying to earn citations in AI answers rather than just rankings. Call +1 321 341 7733, email ab@ab7solutions.com or director@ab7solutions.com, or start at www.ab7solutions.com. We will say plainly if what you need is one good editor rather than a supplier.

Sources: Editorial Freelancers Association, 2026 Editorial Rate Chart (self-reported; 1,100+ respondents, collected November 2025 to mid-January 2026, covering 2025 rates); Google Search Central, Spam policies for Google web search (scaled content abuse, enforcement) and Creating helpful, reliable, people-first content (self-assessment questions, automation disclosure); US Bureau of Labor Statistics, Writers and Authors (median pay, May 2025) and Employer Costs for Employee Compensation (June 2026, released 9 September 2026); Content Marketing Institute and MarketingProfs, B2B Content Marketing Benchmarks, Budgets and Trends (980 B2B marketers, fielded 25 June to 16 August 2024); WebFX, Content Marketing Pricing (vendor-published survey of 350 digital marketers; no fielding date stated).