The solver is not clever and does not need to be. Breadth-first search on a 2,398-node graph is instant, and the guarantee it gives is the product feature.
The solver on the site is described as breadth-first search over the official dictionary, and the phrase "not AI" is doing deliberate work in that description. The distinction matters because it sets the expectation: the tool does not guess at a plausible route. It returns a route that is guaranteed shortest within the accepted word list, or it reports that none exists.
Breadth-first search is the right algorithm because the graph is unweighted and tiny. With roughly two and a half thousand nodes and a bounded branching factor, the search space is measured in thousands of edges, not millions. There is no engineering reason to be clever, and being clever would cost the guarantee. A greedy heuristic would sometimes find a shorter-looking route and sometimes fail, and a solver that is sometimes wrong is worse than no solver at all.
The interesting design question is not the algorithm but the result format. A shortest route is a list of words; the useful part is the first move plus the shape of the rest. That is why the solver's output is a route with the letter changes visible, and why it starts from whatever word the player typed rather than from the daily start. The player is mid-ladder, not at the beginning, and a solver that forces a restart is a solver nobody uses when they are actually stuck.
There is one honest limitation worth repeating: a mathematically legal route is not the same as an accepted route, because the game's private acceptance rules may be stricter than the public list. The tool can only promise within the words it is given, and saying so plainly is better than overselling it.
The solver is worth using as a teacher, not an oracle. Play your attempt first, then check the route on the solver and ask which junction you misjudged. If you want a board to test a new idea from that comparison, click here for the daily round and keep the compared junction in mind.