B2B Content Marketing: What to Publish, and What to Measure First

B2B content marketing is publishing material that helps a buying committee make a decision, in the hope that yours is the vendor they remember and can defend internally. The hard part is not the writing. It is that the person reading will not convert in the session, may never fill in a form, and is one of several people who have to agree.

This is for whoever has to report on a content programme before it has produced pipeline. It is deliberately narrow: the strategy layer, including which bottom-funnel pages to build and how to write for several stakeholders at once, is covered at length in our guide to B2B SEO, and this page does not repeat it. What this page adds is a measurement argument, and a count from our own Search Console that we had not run before writing it.

How is B2B content marketing different from B2C?

Four structural differences, and the second one is the only one that changes what you publish rather than how you phrase it.

	B2C	B2B
Who decides	One person, often immediately	Several people, in different roles, over months
What the content does	Moves someone to buy	Equips an internal champion to argue for you
Volume of demand	Large, shallow	Small, specific, high value per visit
What proves it worked	Conversions in the session	Pipeline, months later, attributed badly

In B2B the reader is frequently not the buyer; they are the person who has to justify the purchase to a finance lead and a security reviewer. Content that only persuades the reader has done half the job. Content that gives the reader something to forward has done all of it.

What should a B2B content programme publish?

Four kinds, and the ratio matters more than the list. Most programmes over-produce the first and never get to the third.

• Answers to the questions buyers already search. The demand exists; you are competing for it. This is where SEO and content overlap and where most of the volume is.
• Objection handlers. Pricing, security, migration, integration, build-versus-buy. Low volume, high intent, and usually the pages a deal actually touches. Which of these to build first is the subject of the B2B SEO guide.
• Material your champion can forward. A comparison they can send, a business case they can adapt, a one-page summary a sceptical colleague will read. Almost nobody makes these.
• Evidence nobody else has. Your own data, your own tests, your own results. The only category a competitor cannot copy.

On the ratio, we have a judgement and no evidence. We would start at roughly half the first kind and split the rest across the other three, and that is a preference rather than a finding: we have not tested it and know of no study that has. Our own publishing rule points the other way, since every post we publish under our three-a-week cap has to carry something only we can say, which is the fourth category. What we would defend is the negative: a programme that is entirely the first kind competes on volume with everyone in the category, which is the most expensive way to compete.

How long before a B2B content programme shows anything?

Longer than any plan allows for, and the useful answer is not a number of months.

We have argued elsewhere that a calendar is the wrong shape for this question, because the timeline is set by where the site is starting from rather than by how long the programme has been running. That holds here. If you want our own read, on a site at this stage the numbers start meaning something at around two quarters, and that is a judgement from one property, not a benchmark.

Our record is a useful calibration because it is unflattering. Over nine weeks from 27 July to 26 September 2026 this blog published 149 posts. Through 29 September they had earned 15 clicks, and 139 of the 149 had earned none. A year's worth of output for most teams, compressed into two months, produced almost nothing measurable in that window. The full account is in what 149 posts earned us, and Google's own guidance on creating helpful, reliable, people-first content treats publishing a lot in the hope that some of it ranks as a warning sign rather than a method.

Where do a content programme's clicks actually come from?

On this site, about half, and the half that came from blog posts cost eighteen times the impressions. We pulled the page-level numbers for this article and the split was sharper than we expected.

Over the 28 days to 27 October 2026, growthhasten.com took 13 clicks from 6,324 impressions at an average position of 57.93. Of those 13 clicks:

Page	Clicks	Impressions	Position
Homepage	2	58	3.33
/tools/hasten-linklens	2	20	4.25
/about	1	27	7.04
The app's home page	1	2	2.50
/tools/content-topic-research	1	1	2.00
Subtotal	7	108
The six blog posts that earned a click	6	1,907	12.64 to 84.79

Seven of the 13 clicks came from 108 impressions, which is 1.7 per cent of everything Google showed. Those five pages are a homepage, an about page, the app's front door and two free tool pages. Not one of them is a blog post. The six blog posts that earned the other six clicks needed 1,907 impressions between them to do it, eighteen times as many for one fewer click.

Two limits on what that shows. The numbers are small enough that a single click moves the picture. And the position column is each page's average across every impression it received, not the position of the impression that was clicked. For those five pages the average is a fair proxy, because each rests on 58 impressions or fewer, and all five sit inside the top eight. For four of the six blog posts it is not a fair proxy: theirs spread across 182 to 1,132 impressions and many queries, which is how two of them average above 80 and still took a click. The other two rest on 11 and 20 impressions, so their averages are as readable as the five above, and they are the two that reached the clickable range: /blog/ai-content-agent at 12.64 and /blog/ai-content-watermarking at 19.15. So this is not evidence that content marketing does not work. It is evidence that on a young site the pages carrying brand and product intent reach the clickable range first, and that a programme reporting on blog impressions would have missed where every one of those seven clicks came from.

What should you measure before the pipeline arrives?

Three leading indicators that can move in a month, plus one that is on the list only so you can take it off your report.

Metric	What it tells you	When it is misleading
Queries in positions 4 to 20	Where a page becomes worth pushing. The clicks themselves mostly land higher	Rarely. This is the honest scoreboard
Clicks per published page	Whether more output is helping	Noisy on any site earning a handful of clicks a month
Organic signups by landing page	Whether the traffic is the right traffic	When the signup is not on the page you ranked
Impressions	Very little on its own	Almost always. Impressions at position 57 are worth nothing

Impressions are on this list to be argued with. Between the 28 days to 27 September 2026 and the 28 days to 25 October 2026, ours fell from 18,302 to 6,213, a drop of two thirds, while average position improved from 70.1 to 55.7. Clicks did not fall with them. At 12 to 18 clicks in a four-week window our click counts are too small to carry an argument on their own, which is why position is the number we read. The drop was good news: Google stopped showing us for queries we were never going to win. A programme reported on impressions would have called that a disaster. We have taken the same argument apart in more detail, on a shorter window, in what an impressions drop actually means.

Should you publish more or improve what exists?

On any site with history, improve what exists. A page with impressions has something to build on; a new page starts at zero.

The test is cheap. Pull your queries at positions 4 to 20 with real impressions. Each one is a page that is close, and moving a close page is usually less work than ranking a new one. Deciding which pages qualify is the subject of auditing the posts you already have, and ordering what comes next is in content planning.

The exception is the fourth content type above. Evidence nobody else has cannot be produced by refreshing, because it depends on running something and writing down what happened, as the table two sections up did.

How does AI search change this?

It raises the cost of generic content and leaves the rest roughly where it was.

Two things we can say from our own work. Crawler access is pass or fail: if an engine's crawler cannot fetch your page, it cannot cite you, and that is testable today. On our own queries, AI Overviews are everywhere: we loaded the twelve queries this site earns the most impressions from on 30 October 2026 and all twelve carried an AI Overview, while none of the twelve did under Google's Web filter. That is one property, one topic area, one day, and it is not a measurement of the web.

What does not follow is that you should restructure everything. We measured all 1,650 H2 sections on this blog against the structural advice and found the corpus already compliant on almost all of it, and still at position 55.7. Structural compliance was not sufficient on its own, which is a weaker claim than it sounds: there is no version of this blog written without that structure and we never measured authority, so the study cannot say what structure contributed either way. The durable response is the same as it was: publish things worth citing, and make sure the engines can read them. Our guide to making a passage quotable separates what is documented from what is inferred.

What does a programme look like when it is working?

Fewer pieces, each tied to a query somebody checked, with a growing count of pages inside the clickable range and a short list of pages the sales team actually sends.

That last clause is the one to act on this week, and it is the one claim on this page with no data behind it, which is precisely why it is worth checking rather than believing. Ask whoever talks to buyers, three of them if you have three and one of them if you are the only one, which of your pages they have pasted into an email in the last month.

None is the answer that tells you the programme is producing for search and nothing for the deal. A list that does not match your top traffic pages is the more common answer and the more useful one: those are the pages to build more of, and they will not be the ones your impressions report puts at the top.