Almost every AI marketing tool is pointed at making more things. The competitive gap in 2026 is on the other side of the line, deciding where a specific asset goes and what it has to become to survive there.
In Content Marketing Institute’s 2026 research, 89% of B2B marketers use AI for content creation and only 38% for social scheduling and posting. The results track that split: 87% say productivity improved, but just 39% say content performance improved and 34% report no change at all. Meanwhile LinkedIn announced in March 2026 that it is reducing reach for recycled thought leadership and posts with video that has nothing to do with the text, and Google’s guidance says the spam trigger is scaled content without added value rather than the use of AI itself. The distribution job worth automating is not producing more variants. It is choosing fewer channels deliberately, adapting properly for each, and fixing the measurement you are optimising against.
Spend 20% of your time creating content and 80% promoting it. You have heard it. It is in every content strategy deck, usually attributed to nobody in particular.
I went looking for the source. It is a blog post from 2013 by one marketer, and the entire evidence base is his own account of publishing 2.54 posts a month and gaining subscribers. No survey, no sample, no control group. The post exists to sell a free video series.
The only actual measurement I could find points the other way. Orbit Media surveyed 211 bloggers and found the average writing time was 3 hours 46 minutes and the average promotion time was 2 hours, so about 47% less time promoting than writing.[18] Small sample, self-selected, 2019, and the authors are honest that their averaging method is rough. But it is real data, and it says people already spend a meaningful chunk of their time on promotion. Not 80%. Not 20% either.
Two more famous numbers in this space also fall apart on inspection, and it is worth knowing because AI will hand you all three within seconds. The claim that 60% to 70% of B2B content goes unused was asserted from a conference stage in 2013 and never written up, and the organisation that made it no longer exists as a domain. The claim that employees have ten times the reach of the company page traces, through a magazine contributor post that names no source, to a 2014 agency infographic launching an employee advocacy product that has since been discontinued.
Content Marketing Institute’s 2026 research, covering 1,015 B2B marketers surveyed between June and August 2025, shows exactly where AI has landed. 89% use AI for content creation. 38% for social scheduling, analysis and automated posting. 36% for email. 16% for advertising optimisation. 12% for predictive targeting.[12]
AI-powered marketing applications B2B marketers use or are actively implementing. Content Marketing Institute and MarketingProfs, 1,015 B2B marketers, fielded June to August 2025.[12]
And the payoff follows the effort. 87% say productivity improved. Only 39% say content performance improved, with 34% reporting no change and 5% saying it got worse. CMI’s own summary is the best line in the whole report: AI helps marketers type faster, not think better.[12]
Sit with the strategic implication for a second. The capability everybody has bought is the crowded one. Ninety per cent of your competitors are generating content faster than they were two years ago, which means faster generation buys you nothing relative to them. The underused capability, using AI to decide where a specific asset goes and what it must become to survive there, is where nobody is competing yet.
It also lines up with what marketers say hurts. CMI’s top three challenges: creating content that prompts a desired action (40%), resource constraints (39%), and measuring content effectiveness (33%).[12] Not one of those is a production-speed problem.
Channel selection is usually done by habit. Here is some evidence to do it with instead.
The Reuters Institute’s 2026 Digital News Report, based on a YouGov survey across 48 markets with around 2,000 respondents each, found that social media and video networks are now ahead of every other source as the most widely used way of getting news globally, at 54%. Television news and publisher websites have both fallen sharply since 2020. By platform: Facebook 43%, YouTube 34%, Instagram 26%, TikTok 20%.[1]
But the finding that should actually shape your plan is this one: X and YouTube are the only two networks where more than half of users say they think of the platform as a useful way of getting news. On Facebook, Instagram and TikTok, most people say they generally only see news when they are online for other reasons.[1]
That is a different job for your content on each platform. On YouTube you can answer a question someone came looking for. On Instagram you have to earn attention from someone who came for something else entirely. Feeding the same adapted asset into both and expecting comparable results is the mistake, and it is precisely the mistake that one-click multi-channel AI adaptation encourages.
A caution on the numbers, because I would rather you had it. Reuters uses online panels, states plainly that a conventional margin of error cannot be computed, and warns that online samples under-represent older and less affluent people. LinkedIn is not measured in the report at all, so do not let anyone imply that it is.[1]
Where does that leave organic search? Under real pressure. Chartbeat data reported by Reuters shows Google organic traffic to more than 2,500 sites fell 33% globally and 38% in the US between November 2024 and November 2025.[1] That is a news publisher panel rather than B2B blogs, so read it as a direction rather than your number.
And the baseline nobody enjoys: Ahrefs found in 2023 that 96.55% of pages in its roughly 14-billion-page index got zero traffic from Google, with a further 1.94% getting between one and ten monthly visits. To their credit they publish the caveats themselves, noting the sample is biased toward the quality side of the web and that their traffic figures are estimates.[3] Publishing without a distribution plan is, statistically, publishing into nothing. We have covered the search side of this separately in zero-click search strategy.
You do not have to guess at this. Both LinkedIn and Google published guidance in the past year that speaks directly to AI-assisted distribution, and the guidance is more specific than most of the commentary about it.
Tim Jurka, writing on LinkedIn’s own channel, announced that the feed would show less generic content and less engagement bait, naming three things specifically: less content that says “Comment yes if you agree”; less posts featuring a video that has nothing to do with the associated text post, which he says should no longer gain additional reach; and less recycled thought leadership that does not say much in terms of substance or insight.[4] The same post commits to making engagement pods ineffective and curbing comment automation.
Read that middle item again if your workflow attaches a generic clip to every post. That is now named as a demotion signal.
LinkedIn’s engineering blog explains the underlying mechanism, which is dwell time rather than clicks. Their own words: clicks are noisy indicators of engagement, because a member may click an article and close it within seconds, which they call click bounces. And skipping is treated as a negative outcome, with post scores reduced in proportion to predicted skip probability.[5] Note that LinkedIn has never published the skip threshold in seconds, so any specific figure you have seen quoted was invented by somebody.
The most operationally useful thing LinkedIn publishes is buried in a help article about suggested posts, which are the ones shown to people outside your network. LinkedIn says it tries to serve only content that is valuable and relevant to that wider audience, giving as examples ensuring the post provides context for the topic being discussed, is not overly promotional, and has all its aspects, text, images and video, relating to one another.[6] It is phrased as intent rather than rule, but it is still the closest thing to a checklist they have published.
Google’s position is that provenance is not the test. Their documentation states that using generative AI tools to generate many pages without adding value for users may violate the spam policy on scaled content abuse.[7] Using AI is fine. Using it to make volume is not.
Their generative AI optimisation guidance is unusually blunt about the tactics being sold right now. On special files: you do not need to create new machine-readable files, AI text files, markup or Markdown to appear in Google Search including its generative features, because Google Search does not use them. On outreach: seeking inauthentic mentions across the web is not as helpful as it might seem. And on producing content for every query variation, doing so primarily to manipulate rankings or generative AI responses violates the scaled content abuse policy.[8]
Google also draws a distinction worth stealing for your editorial meetings. Commodity content, their example being “7 Tips for First-Time Homebuyers”, is based on common knowledge. Non-commodity content, their example being “Why We Waived the Inspection and Saved Money”, provides a unique expert or experienced take.[8] AI is excellent at the first kind and cannot produce the second without you. There is more on this in our answer engine optimisation guide and in getting your brand cited in ChatGPT.
Meta has twice told Pages on the record that reach would fall, in 2014 and 2018, and has never quantified it. What its own Widely Viewed Content Report shows is the structural shift: content from friends fell from 57.0% of US Facebook Feed views in Q2 2021 to 20.1% in Q4 2025, while unconnected recommended content rose from 8.0% to 41.0%.[9] Read it as a direction rather than a precise comparison, because Meta renamed and redefined those categories between editions and the report page only ever serves the current quarter. The direction is unambiguous enough: being recommended now matters far more than being followed.
The standard AI distribution workflow is: paste article, ask for fifteen channel variants, schedule. Given everything above, that workflow is now actively working against you on two major platforms simultaneously.
Here is what I do instead, and it takes about the same amount of time.
Use the evidence on what people go to each platform for.[1] For most B2B teams that means LinkedIn, email and one place where your specific audience actually congregates. CMI’s respondents named LinkedIn (76%), email newsletters (54%) and speaking events or webinars (52%) as the most effective channels for thought leadership specifically.[12] That is perceived effectiveness for one content type, not a general ranking, so treat it as a starting point.
Most channel adaptation prompts ask for a summary, which produces exactly the recycled thought leadership LinkedIn says it is demoting.[4] Ask instead: “What is the one claim in this piece that a knowledgeable person might disagree with, and what is the evidence for it?” Then build the post around that. It gives a reader a reason to stop, which is what dwell time actually measures.[5]
If you attach a video, it has to be about the thing the text is about. LinkedIn named the mismatch explicitly.[4] This sounds obvious and I see it broken weekly by teams using bulk generation tools that pair posts with stock footage.
It is the only part most people read before deciding to skip, and skipping is the signal that costs you.[5] AI openings are recognisable, and the recognisable version is the demoted version.
For the mechanics of adapting a single asset properly, we have written up turning one post into ten pieces and writing LinkedIn posts with AI.
Every scheduling vendor publishes a best-time-to-post chart. They contradict each other flatly, which should tell you something before you read another one.
For LinkedIn: Sprout says Tuesday to Thursday, 11am to 5pm. Hootsuite says Tuesday and Wednesday, 4am to 6am. Buffer, analysing 4.8 million LinkedIn posts, says Wednesday 4pm, that Monday and Tuesday see the lowest engagement, and that traditional morning hours now tend to see lower engagement. So one vendor’s best window is another’s worst.
The largest peer-reviewed test of the underlying idea is worth more than all of them. Researchers analysed over a billion reactions across hundreds of millions of messages, validating out of sample on 500,000 active users and more than 25 million messages. Generic schedules determined per timezone and not personalised per user produced, in their words, little to no increase in reactions received, and on Twitter often left users below their own baseline. The frequently quoted 17% figure was an upper bound, applying only to a fully personalised schedule computed from each account’s own first-degree audience and weighted by individual reaction propensity, and even that dropped to 4% on Twitter.[16]
In other words, the published charts are precisely the thing the research says does not work, and the thing that does work cannot be published as a chart because it is different for every account.
| Tool | Basis for the recommendation | What to know |
|---|---|---|
| Sprout Social | 16 weeks of your own engagement per profile, refreshed weekly, scored in 5-minute increments | Genuinely your data. Falls back to global trends if a profile lacks engagement[14] |
| Hootsuite | Your last 30 days of results, plus when your followers are online | Shows industry standard for your timezone until it has your data[15] |
| Buffer | Platform-wide benchmark from active US-timezone accounts, z-scored | The marketing line says “when your audience has been most active”; the FAQ on the same page says otherwise[13] |
Taken from each vendor's own support documentation. Only one of the three is computing from your specific audience by default.
Buffer’s own page contains both claims, which I find genuinely useful as a teaching example: the promotional sentence says the times are based on when your audience has been most active, and the FAQ further down says the schedule is configured by analysing performance across active real accounts in US timezones.[13] To their credit they also say to treat the times as a starting point rather than a guarantee.
Practical answer: export your own last 90 days and look. That is the personalised signal the research says is the only one that works, and you already own it. Notably, LinkedIn itself publishes no best-time-to-post recommendation anywhere.
This is the part I would fix first, because an AI agent optimising your distribution will do so faster and more confidently than a human, and right now three of the signals it would optimise against are corrupted.
Search referral. Pew Research Center tracked the browsing of 900 US adults across every URL they visited in March 2025, covering 68,879 unique Google searches. When an AI summary appeared, users clicked a traditional search result on 8% of visits, against 15% when no summary appeared. Clicks on links inside the summary itself happened on 1% of visits. Sessions ended after 26% of pages with a summary versus 16% without.[2]
Two honest caveats Pew flags itself: the summaries were re-scraped weeks after the browsing, and queries that trigger summaries are longer and more question-shaped, so intent is confounded with summary presence. Also worth knowing for planning: only 18% of searches produced a summary at all, rising from 8% of one-to-two-word searches to 53% of searches with ten or more words.[2] The exposure is concentrated in long, question-shaped queries.
Email opens. Broken since Apple introduced Mail Privacy Protection. Publishers openly attribute their own open-rate increases to auto-opens, and the two best-documented benchmark sets in the market report all-industry open rates around 21.5% and 41.24% for the same metric. That twenty-point gap is measurement, not reader behaviour. Use click rate and click-to-open, and treat opens as directional at best.
Channel attribution. Google documents Direct as traffic that does not have a clear referral source, and in GA4 anything it cannot classify at all lands in Unassigned rather than Direct.[10] The practical causes marketers run into are mostly untagged links, redirects and shorteners that drop parameters along the way, and links opened from PDFs, decks and messaging apps. Google does not publish that list, so treat it as practitioner experience rather than documentation. The fix is the same either way, and it is the next paragraph.
Do that before you let anything automated optimise on your behalf. Otherwise you are pointing a very fast optimiser at three broken instruments. If attribution is where your reporting hurts most, using AI for marketing attribution goes deeper.
You scale variants instead of choosing channels. This is the big one, and it is now explicitly a demotion signal on LinkedIn and a spam-policy trigger on Google.[4,7] Fifteen adapted posts is not fifteen times the distribution. It is one idea, diluted, in fourteen places where it does not fit.
You chase mentions because a vendor told you to. Google’s own words: seeking inauthentic mentions across the web is not as helpful as it might seem, and you do not need special AI files or markup to appear in its generative features because Search does not use them.[8] A lot of money is currently being spent on both.
You take the chart over your own export. The vendors disagree with each other, the peer-reviewed evidence says generic timezone schedules deliver little to nothing, and one of the three tools above is not using your audience data at all by default.[13,16]
You measure with instruments you have not checked. Search referral is falling, opens are inflated by auto-opens, and traffic you failed to tag lands somewhere that tells you nothing.[2,10]
You confuse faster with better. 87% productivity improvement, 39% performance improvement, 34% no change at all.[12] If your content output has doubled and your pipeline has not moved, that gap is the whole story and more volume will not close it.
One thing to do this week: pick your three channels, write down what people actually come to each one for, and delete the other six from your plan. Everything in this article gets easier once that list is short. And if you want the strategic layer above it, we have building a content marketing strategy with AI.
There is no evidence behind that rule. It comes from a 2013 blog post whose entire evidence base is the author’s own subscriber count, published as a lead magnet. The only measurement I could verify points the other way: a survey of 211 bloggers found average writing time of 3 hours 46 minutes against 2 hours of promotion, which is 47% less time promoting than writing. Treat the split as a judgement call about your own situation, not a law.
Not for being AI-generated. LinkedIn’s March 2026 announcement names three specific things it is reducing reach for: engagement bait such as asking people to comment yes, posts featuring video that has nothing to do with the associated text, and recycled thought leadership that lacks substance or insight. Google’s position is structurally the same, targeting scaled content without added value rather than the use of AI. The test on both platforms is quality and coherence, not provenance.
Whatever your own last 90 days say, and no published chart can tell you. The largest peer-reviewed study of this, covering over a billion reactions across hundreds of millions of messages, found that generic schedules set per timezone and not personalised produced little to no increase in reactions, and on Twitter often performed below the user’s own baseline. The 17% figure frequently quoted from that study was an upper bound that applied only to fully personalised schedules built from an individual account’s own audience. The vendor charts also contradict each other, with one naming as best the exact window another names as worst.
Check the documentation rather than the marketing copy. Sprout Social’s ViralPost analyses 16 weeks of your own engagement per profile, refreshed weekly and scored in five-minute increments. Hootsuite uses your last 30 days, falling back to an industry standard until it has your data. Buffer’s FAQ states that its recommended times come from analysing performance across active real accounts in US timezones, even though a sentence higher up the same page describes them as based on when your audience has been most active.
It is reducing clicks and making measurement harder. Pew Research Center tracked 900 US adults across 68,879 Google searches in March 2025 and found users clicked a traditional result on 8% of visits when an AI summary was present, against 15% when it was not, with clicks into the summary’s own links happening on 1% of visits. Only 18% of searches produced a summary, though that rose to 53% for searches of ten words or more. Chartbeat data reported by Reuters Institute shows Google organic traffic to over 2,500 news publisher sites fell 33% globally between November 2024 and November 2025, which is a direction to plan around rather than a number that will match your own analytics.
This article deliberately excludes several of the most-quoted statistics in content distribution because they could not be traced to any study. The 80/20 promotion rule comes from one blogger’s 2013 lead-magnet post. The claim that 60% to 70% of B2B content goes unused was asserted from a conference stage in 2013 and never published. The claim that employees have ten times the reach of a company page traces to a 2014 agency infographic promoting a since-discontinued product. What is here instead comes from Reuters Institute and Pew Research Center survey and browsing data, platform-official documentation from LinkedIn, Google and Meta, vendor support documentation for scheduling tools, peer-reviewed research on posting time, and Content Marketing Institute’s annual B2B survey.
Weekly tips. Real examples. Practical help for busy professionals.
We care about your data, check out our privacy policy.