Before this category existed, a stuck student had a textbook answer at the back of the book and no route from the problem to it. Photomath's contribution is that it shows the route. Point a camera at an equation and get the steps, with the option to expand any step you do not follow. That is a real improvement on both the back of the book and the answer sharing sites that preceded it, and offering several different methods for the same problem is better still, because seeing that a quadratic can be factored, completed or formula solved teaches something about mathematics that a single answer never does.
Handwriting recognition works well enough that it handles a page of your own working, which matters because most homework is not printed. The free tier covers the actual solving rather than showing you an answer and charging for the explanation, which would have been the obvious and cynical structure. Credit where it is due for not doing that. The pedagogical objection is the obvious one and it is not a reason to dismiss the tool, only to be honest about it.
A student can photograph twenty problems, copy twenty sets of steps, hand in completed homework and have learned nothing. The design makes this trivially easy and the material does not engage with it. The counterargument that the steps are visible so the student might read them is true and it is a hope rather than a mechanism, and there is no feature and no guidance that nudges toward engagement over transcription. Something as simple as hiding the next step until the student attempts it would change the incentive, and its absence is telling.
The explanations themselves are procedural. They tell you to move a term, apply a rule, factor an expression, and they are considerably lighter on why that operation is legitimate or what it means. For a student who understands the concept and lost the thread, that is fine and sufficient. For a student who does not understand why you can do the same thing to both sides, procedural steps produce successful mimicry and no comprehension, and mimicry collapses the moment the problem is phrased differently.
Word problems expose this immediately, and they are also where the app is weakest, because translating a described situation into mathematics is the actual skill being taught and the part the camera cannot help with. What I would most want and cannot find is anything written for the adults in the situation. Parents want to know whether to allow it, teachers want to know how to set homework in a world where it exists, and both would benefit from practical guidance, such as using it to check completed work rather than to produce it, or asking a student to explain a step back. That material would cost little to write and would improve outcomes more than any feature.
Being part of Google now, the app is stable and well maintained and the direction is not obviously in the company's hands as an independent product. Three point three. A well built tool that made the right core decision about showing working, sitting inside an ecosystem that has thought carefully about recognising handwriting and not at all about whether anybody is learning.