HubSpot's blog went AI-first: same post count, about 2 fewer hours each
HubSpot's blog manager says a year of AI-first writing cut a post from about 10.25 hours to 8.25, with output flat and fact-checking added as a new step.
At a glance
- Company
- HubSpot (blog writing team) · Technology and software · Size not stated
- Team and use case
- Marketing · Content production
- Tools
- Claude, ChatGPT, NotebookLM, Custom quality checker in Google Docs (built in-house)
- Stage
- In production, one team, since June 2025
- Results
- Hours per standard blog post: About 10.25 (range 9 to 11) → About 8.25 (range 7 to 9)
- Posts published: Not stated → About the same
- Fact-checking time per post: A light final step → An estimated 30 to 60 minutes
- Cost
- Not disclosed. No licence, usage or build figure; the manager built the quality checker.
- People
- Writers chose their own AI methods; no job cuts or hires are mentioned. Some felt pressure to write faster and more until workload was recalibrated. Fact-checking is a new dedicated step (who does it is not stated), and a required Google Docs check flags quality problems. Team size not disclosed.
- Evidence
- Company claim, no method
The company states a result without saying how it was measured.
What this leaves out: Averages from HubSpot's own time tracking, shown in charts with no number of posts, no period and no per-stage hours after AI; the fact-checking figure is the manager's estimate. No traffic or lead result is given, and HubSpot sells marketing software.
HubSpot's blog writing team moved to what it calls an AI-first writing model in June 2025. A standard post now takes about 8.25 hours against about 10.25 before, and the team publishes about the same number of posts, according to [an account by Jamie Juviler](https://blog.hubspot.com/marketing/how-ai-changed-our-editorial-process), who manages the team, published on HubSpot's blog on 19 October 2026.
That is 2 hours a post, or about 20% (2 ÷ 10.25). Drafting alone took five to seven hours before AI, so most of what AI saved in drafting went somewhere else.
Where the time went
- Longer posts, not more of them. A couple of months in, time-tracking data and conversations with writers showed post count flat while posts became longer and more detailed. Writers spent the saved time testing software, sourcing expert quotes and reviewing tools.
- Fact-checking. Juviler reports fabricated statistics, misattributed quotes and wrong terminology in AI output. Checking facts went from a light final pass to its own phase, estimated at 30 to 60 minutes for a typical post and longer for anything with statistics, quotes or product details. In some cases AI drafting took longer than writing by hand.
- Competitor research. On a listicle of competitor products, a writer found the pricing the AI supplied was fully out of date. Research on competitor pricing and features takes roughly as long as before; general topic research got faster.
- Quality checks. AI drafts brought stock phrases, repeated sentences, text lifted from competing pages, broken links and missing image descriptions. Juviler built a checker inside Google Docs that flags these, and running it is now required.
AI hadn’t decreased the workload. In many ways, it had increased the workload.
Jamie Juviler, manager of HubSpot's blog writing team, HubSpot blog, 19 October 2026
The people side
Writers were given Claude, ChatGPT and NotebookLM and left to work out their own methods; the team's rule was experimentation first, structure later. Juviler writes that some writers felt two expectations at once, to write faster and to write more, and that the workload had to be recalibrated. Writers who came to the job for the craft found the change harder than those who care most about results, and the ones who pushed back on early workflows showed what needed fixing. The post mentions no job cuts or hires and does not give the size of the team. It does not say whether the writer or an editor does the new fact-checking step.
Reading this at work? Get one story like it explained every morning.
A five-minute read every morning at 7am ET: one story explained, two in brief, what else happened, one number. Unsubscribe in one click.
The catch
The hour totals are averages from HubSpot's internal tracking, shown in two charts without the number of posts, the period, or hours for each stage after AI. The fact-checking figure is Juviler's estimate. The post gives no licence cost and no traffic, lead or search-ranking result for the longer posts, so it shows the time and not whether the posts perform better. HubSpot sells marketing software, and the page carries adverts for its own AI agent builder; the tools the team describes using are other companies'.
What to do if you run content or marketing operations
- Set the target for AI-assisted writing in hours per post and what each post must contain. HubSpot's post count did not move in the first year.
- Add fact-checking as a named step with an owner, and budget 30 to 60 minutes a post for it until your own tracking says otherwise.
- Keep competitor pricing and feature research manual, and have a person read the vendor's live page before anything about a competitor is published.
Take this to work
Forward this
HubSpot's blog team says a year of AI-first writing took a standard post from about 10.25 hours to 8.25 while post count stayed about the same, with fact-checking added as a 30 to 60 minute step, according to its manager's account of 19 October 2026. A content plan that counts on AI for more posts should be tested against that: the reported saving is about 20% of hours, and it went into longer posts.
Questions to ask
- How many hours does a post take us today, by stage, and who is tracking it?
- Who fact-checks AI-assisted drafts, how long does it take, and what share of drafts had an error?
- Did the longer posts earn more traffic or leads than the old ones? HubSpot's account does not say.
- What do the AI tools cost per writer per month, and is that more or less than the hours saved?
- Which content types do we keep manual, such as competitor comparisons and original research?
The arithmetic
Writer hours freed per month = posts per month × (hours per post before − hours per post now). Your inputs: posts per month and tracked hours per post before and after.
HubSpot gives about 10.25 hours before and about 8.25 now: 10.25 − 8.25 = 2 hours a post, and 2 ÷ 10.25 = about 20%. It does not give posts per month, team size or tool cost, so the monthly total and the return cannot be worked out from the post.
Questions this article answers
How much time did AI save HubSpot's blog writers?
About 2 hours a post. HubSpot's blog manager reports a standard post takes about 8.25 hours with AI against about 10.25 before, on averages from the team's internal time tracking published on 19 October 2026.
Did HubSpot publish more blog posts after adopting AI?
No. The manager writes that the quantity of posts stayed about the same while posts became longer, more detailed and more demanding to produce.
How long does fact-checking AI-written content take?
HubSpot's blog manager estimates a dedicated fact-checking step adds 30 to 60 minutes to a typical post, and more for pieces with statistics, quotes or product information.
Which AI tools does HubSpot's blog team use?
Claude, ChatGPT and NotebookLM, plus a quality checker the manager built inside Google Docs, according to the 19 October 2026 post. The cost is not disclosed.
Adecco: applicants its AI agent prescreens are 2x as likely to be hired
Adecco's recruiters start from a shortlist an AI agent scores in chat. Applicants opt in, so the twice-as-likely-to-be-hired figure needs a control group.
Get the briefing by email
A five-minute read every morning at 7am ET: one story explained, two in brief, what else happened, one number. Unsubscribe in one click.