The interesting question about GLP-1 pricing is not what the cheapest seller charges. It is how often the number on the page is the number on the statement. Across this desk’s 962 published price rows, 182 carry a correction — 19% of everything recorded.
How big the gap is
The median corrected row understates by $39.50 a month. That is not a rounding error and it is not a scandal either; it is roughly the cost of a membership most sellers describe openly somewhere on the site — small beside the spread shown in the full price distribution. The largest single gap in the corpus is $233 a month, which over six months is $1,398 that the headline does not mention.
Four things cause almost all of it
A membership charged beside the medication. The commonest by some way. The drug is priced honestly and a standing fee sits next to it, often on a different page.
A program or care fee. The same arrangement under a different noun. Some sellers fold it in and publish the combined figure, which is the behavior worth rewarding.
A consultation billed separately. Usually one-off rather than recurring, which makes it smaller but no less real on a first bill.
A month that is four weeks. The quietest of the four. Thirteen fills a year rather than twelve is about an eight per cent premium, invisible unless a reader reads the interval rather than the number — like the format premium, which also hides in plain sight.
What this means for comparing
Comparing two advertised figures is only meaningful when both are whole. Roughly one row in five here is not, so a like-for-like comparison has to start by asking what each number excludes — and the answer is usually published, just not on the page carrying the price. The term and prepay check lists every seller whose own note names a condition attached to its rate.
Each figure here is recalculated when the page builds, from prices this desk read off the sellers’ own pages and stamped with the day it read them. The newest stamp is September 2026. None of it is bought in, estimated, or carried over from somebody else’s dataset.