I feel like there are a few unstated assumptions in whsheet's post. Even with those, of course, no advice is universal or without exception. They do make a good point, though.

My version of the thought process is that, all else equal, SEO has three key disadvantages:

1. Attribution is often harder than some other channels, and your experimental design has to account for that.

2. There is a less direct, unitary through-line from dollars invested to dollars earned. Say you spend X dollars on various SEO improvements this month, and you get Y units of growth (conversions, revenue, etc.) that you can attribute to that SEO work. It is less reliable to assume (relative to other channels) that spending 10X on SEO improvements next month will reward you with 10Y units of growth.

3. If SEO is your dominant channel at 95%, and you ever hit the maturity in the channel, the effect of the above is can be fatal to a growth-oriented startup. However you measure your SEO-based CAC, it is probably going to be lower than the CAC of whatever your backup/next channel ends up being, and lower by a large amount. If you have to add in a lot of SEM with your SEO, CAC goes way up, and if your LTV was too precariously close to your CAC already, you are very likely to go negative on your unit economics.

#2 and #3 are what I assumed whsheet was intending as I read their post.