Production Speed Was Never The Constraint
Google Ads US demand for "elearning authoring tool" fell from about 8100 monthly searches in September 2025 to about 3600 in June 2026. Over the same window, "scorm authoring tool" cooled from about 320 to about 140. We pulled that history from the DataForSEO Labs Google Ads US keyword database this week.
Meanwhile, ATD's 2025 State of the Industry put average cost per learning hour used at $165 for 2024, up from $123 the year before, a 34% increase, while formal learning hours per employee fell to 13.7 from 17.4. Teams are paying more per hour of training delivered even as search interest in the tools used to build that training cools.
The Economics Inverted Before The Tooling Did
Read the ATD numbers together, and the story is not a budget cut. Average direct expenditure per employee was $1,054 in 2024, down only slightly, and organizations put 2.9% of revenue into talent development, the highest ratio in 5 years. Money did not leave. Delivered learning did. Fewer formal hours reached each employee, so the same spend divided into a smaller denominator, and cost per learning hour used jumped by a third.
That is worth sitting with, because cost per learning hour used is the rare L&D metric that refuses to reward production for its own sake. It counts hours consumed, not hours built. A team can double its catalog and make that number worse.
Cooling Search Demand Is Not Cooling Need
Now put the search data next to it. The decline in "elearning authoring tool" and "scorm authoring tool" queries is steep, and it is tempting to read it as L&D losing interest in authoring altogether. A more careful reading is that fewer people are shopping the category the old way, comparing named desktop tools against each other, while the work itself continues.
The delivery side supports that. "Learning management system" fell from about 33,100 monthly searches in September 2025 to about 9900 in June 2026, and even after that decline it remains, by a wide margin, the highest-volume term of the three. Buyers may be searching less, but they have not stopped needing courses to land somewhere that tracks completion.
Where AI Authoring Actually Changes The Math
If cost per learning hour used is the number under pressure, AI authoring helps in one specific way: by producing hours that actually get consumed. Compressing production time lowers the numerator. It does nothing for the denominator on its own, and it can quietly hurt it. A generated course that learners abandon halfway still cost something to make, and it delivered fewer used hours than the shorter course nobody got around to building.
So the case for AI authoring in 2026 is not "we can make more." It is "we can afford to make the right thing, and remake it when it stops working." Cheap iteration is the actual gift here. Cheap volume is a trap, and cost per learning hour used is the metric that will find it.
What To Check Before You Trust A Generation Demo
Speed demos are common and most of them look alike. Three questions separate them.
- Does the tool force outcome-first structure before slides, or does it mirror the source document? A generator pointed at a policy PDF will happily hand back a course shaped like a policy PDF. Decide what the learner should be able to do differently, then watch whether the tool ever asks that question itself.
- Are knowledge checks gates, or optional decoration? A question a learner can click past is not assessment. If nothing blocks progress until understanding is demonstrated, your completion data is describing attendance.
- Can you export SCORM 1.2 and SCORM 2004, and re-dispatch updates when the same course must live in more than one LMS? ATD found the three most common content areas for training were new-employee orientation, mandatory and compliance training, and managerial and supervisory training. Two of those three get audited. Auditable delivery is not a nice-to-have in that context, and re-uploading a corrected package by hand across several systems is where multi-LMS good intentions usually die.
The Squeeze Nobody Budgeted For
The same ATD research found that 55% of organizations delivered AI technical skills training in 2024, and that 75% expected to increase AI spending in the following fiscal year. L&D teams are being asked to teach AI to everyone else and to adopt it themselves, in the same period, while hours per employee keep falling and each of those hours costs more to deliver.
That is a genuine squeeze, and it is exactly why the evaluation questions matter more than the demo. A team under that much deadline pressure will reach for whatever ships. The only durable protection is choosing tooling where the well-designed version is also the fastest version to produce.
The point is not that AI authoring is optional. It is that cheaper, faster production only helps if cost-per-hour pressure and cooler tool-shopping demand push teams toward workflows that keep pedagogy and delivery standards intact by default.
Sources And Methodology
Every figure above comes from one of two places. Both are listed here so readers can check them independently.
Training Economics
ATD, 2025 State of the Industry: Talent Development Benchmarks and Trends, based on a survey of 539 talent development professionals. Figures cited: average cost per learning hour used of $165 in 2024 against $123 in 2023, a 34% increase; average formal learning hours per employee of 13.7 against 17.4; average direct expenditure per employee of $1,054; 2.9% of revenue invested in talent development; 55% of organizations providing AI technical skills training in 2024; 75% expecting to increase AI spending the following fiscal year; and the 3 most common content areas for training being new-employee orientation, mandatory and compliance training, and managerial and supervisory training.
- Full report: 2025 State of the Industry.
Search Demand
Monthly search volumes were retrieved on 4 August 2026 from the DataForSEO Labs historical keyword data endpoint, which draws Google Ads keyword metrics for a specified location and language and refreshes monthly. Location: United States. Language: English. The values cited are the monthly series for September 2025 and June 2026: "elearning authoring tool" about 8100 falling to about 3600; "scorm authoring tool" about 320 falling to about 140; "learning management system" about 33,100 falling to about 9900.
- Endpoint documentation: dataforseo_labs/google/historical_keyword_data.
One Methodological Caveat, Stated Plainly
A keyword's monthly search-volume series and its headline search-volume figure are two different measurements, and year-over-year trend percentages are usually calculated on the second. Every number quoted in this article is from the monthly series. Anyone reproducing this analysis should compare like with like, because mixing the two produces percentages that look authoritative and describe nothing.
