Granted, not everyone might see them as good, and a lot of people's opinion probably comes from other people talking about them rather than experimenting with them in a real game.
Without going into details, and save for the few early levels, during which you might have seen a few skunks being conjured to great effect, a top-level summoning slot brings up a creature between 4 and 5 levels below the party.
Due to how encounter math works, a creature of this level is counted as between 0 and 10 XP in the rules for building an encounter as its chances to hit are too low to matter against the player (-5 to hit against +5 to all defense at a minimum, often more from proficiency upgrades).
Of course, that's for abilities targeting defenses, surely I just have to pick things that don't target defenses or satisfy myself with spawning an annoying flanking/body blocking buddy? This is correct, some very select support-oriented monsters, like the Satyr or, in an undead campaign, the Deathless Acolyte can give an amazing boost for their level in a vacuum; but that's before considering what truly seals this pan of the game for me
It's woefully action intensive for the caster. A good way to see it is to say that you're spending 3 actions to slow 1 yourself in order to add a level -5, stunned 1 monster on your side of the board, and if the support action of a Satyr might feel pretty good, is it really compared to other uses of 1 action for the caster, like using a composition cantrip, an appropriate metamagic, or using a well-chosen skill action like bon mot or demoralize? and that's excluding the initial 3 action opportunity cost you could have spent on a more potent spell
In short, there is a reason why level -5 creatures don't count in the encounter budget, and while a well-chosen one might impact the fight positively, 2 of its actions are almost never going to be better than 1 action of a creature 5 levels higher;
Of course, that doesn't mean the spell is useless, out of combat in the blood lord adventure, for example, a single cast at 4th level of animate dead can be used by the Wizard to heal everyone for 20 + 3 x (2d8+16) to distribute on the most injured in a minute with a deathless acolyte; that's amazing, and notably way more than the 0 a wizard would be able to provide otherwise. Similarly, if you know something is booby-trapped and you don't want to risk your rogue, a Crawling Hand will happily eat and "disarm" it for the party for the cheap price of a 1st level spell.
Summoning was specifically defanged in combat, probably as a design concern about minion spam that was prevalent in previous editions, so just... don't use it in combat and demoralize/bon mot every turn instead, you'll be doing more good for your party
I think the designers were worried about summoning becoming too good as the pool of available creatures expanded with every bestiary published. I feel like that was a real issue in 1st edition. Unfortunately, I think they went too far in trying to avoid that problem. I wish they had gone the same route as the various combat form spells and just provided a couple of base templates for summons, with easily applied modifiers as the spells were heightened. It would take some variety away, but it would hopefully allow summons to be a little more useful than "I'm here to absorb a single hit at -10 MAP, then die."
I don't want summons to be too good in combat -- they shouldn't take the spotlight from martial characters, or pet-based classes, after all. At the moment, though, I feel like they're tuned too far in the other direction.
Yup, summoning being, let's be real, a pretty bad combat option is likely by design, and probably not a bad design decision if anything, they did make a class centered around the concept if that's your thing (summoner), with a reskinned animal companion archetype for the "necromancer" trope, while ensuring people couldn't cast "conjure barbarian army" and make a single fight last 4h during which it's your turn for 80% of it.
I don't think Paizo wants summoning spells to be used, but a lot of people online will swear it's a good option. It really isn't and clearly wasn't meant to be, and that's OK. I usually consider that PL-2 to PL-3 is guaranteed to be weaker than a single character action for action (even including the 1 sustain for 2 actions) and would have been a relatively "safe" tuning point to use, PL-5 is almost doubling the safety margin, that plus the fact Paizo happily erratas outstanding summon options that would be viable despite the level disparity is a clear indication they wanted to make real sure this wouldn't be a strategy