The "low competition" score in every keyword tool is lying to you, and here is the proof

I pulled 600 keywords that three popular tools all scored as low difficulty — under 20 on their respective scales — and checked what was actually ranking.

What I found in the top 10
At least one site with 500k+ monthly traffic: 71% of keywords
A forum or Reddit thread in the top 5: 44%
A result older than four years with no updates: 38%
An actual small site under 10k traffic in the top 3: 9%

So nine percent of "low competition" keywords had a small site visibly winning. That is the number that matters and no tool reports it.

Why the scores are wrong
Every difficulty score is a function of backlinks to the ranking pages. That made sense in 2015. It does not describe a SERP where a 200-authority brand page with eleven links outranks a perfect 60-link niche page because of everything else the domain has earned.

What I do instead
I open the SERP. That is it. I look for at least two results that are weaker than what I can make — a thin page, a forum thread, something stale, an irrelevant result the engine clearly settled for. If I cannot find two, I skip the keyword no matter what the tool says.

It is slower and it does not scale into a spreadsheet of 10,000 rows, which is exactly why it still works.

The 9% number is going in my notes. I have been running the manual SERP check for two years and could never justify it to clients who wanted a difficulty column. Now I can.

Counterpoint, partially: the scores are not useless, they are just a filter not a decision. I use them to cut 10,000 keywords to 400, then eyeball all 400. Throwing out the tool entirely costs you reach.

That is a reasonable position and probably the right one at scale. My objection is to people who stop at the score, which from the content I see is most people.

Did you control for intent? A lot of low-difficulty keywords are low difficulty because nobody searching them wants to buy anything, and the big site ranking there is not trying.

Not rigorously, no. Good catch. Informational long tail was overrepresented in the sample, which probably inflates the "big site in top 10" figure.

The "find two results weaker than what you can make" heuristic is the most portable thing in this thread. That is a writing brief as much as a keyword filter.