Let's consider a possible scenario, with a simplified assumption that red hair is a simple Mendelian recessive trait.
Let's say we have a population that is 10% red hair. Everyone with red hair lives on one island, and they only breed among each other. They all have red hair, so we can infer they are all homozygous recessive (rr) for that trait. Everyone without red hair lives on the other island. Red hair never appears in this population, so we can infer they are all homozygous dominant for the non-red hair trait (RR).
Now, we develop inter-island transportation, and breeding becomes randomly assorted among the whole population. The allele frequencies stay the same: 10% r and 90% R. There is no selection based on the red hair trait. However, from Hardy-Weinberg we would calculate that only q^2 = 0.1^2 = 1% are homozygous recessive. 2 * p * q = 2 * 0.9 * 0.1 = 18% are heterozygous carriers for red hair, but they don't express the phenotype.
Just by changing the mating scheme, you can go from 10% with red hair to 1% without any change in allele frequencies.
It seems reasonable to infer that something like this is happening in the real world. People with red hair are much more common in certain places in the world. When people tend to mate only with people in their local communities, the likelihood of homozygous recessive individuals is higher than when people immigrate and mate with partners around the world. That doesn't mean there is any selection for the red hair allele, just a consequence of the mating scheme. That's the takeaway from the National Geographic quote you found on Wikipedia:
"while redheads may decline, the potential for red isn't going away"
In this statement, "redheads" are the homozygous recessive folks that express the red hair phenotype. "Potential for red" is a colloquial way of referring to the allele frequency, "r".
Of course you could also get actual changes in allele frequency for reasons that have nothing to do with fitness of the red hair phenotype. If you think about population growth on a regional basis, if there is more population growth in countries with lower frequency of a red hair allele, then the frequency of red hair in the human population as a whole would decrease. It seems likely this is occurring today since regions with higher frequency of red hair (northern Europe/UK/Ireland, also the US) have lower population growth than those with lower frequency (Asia, Africa, South America). There's no hint that this would result in a disappearance of the trait, though.