The short version

The shoulder is four possible targets, not one. Subacromial space, the joint itself, the AC joint, the biceps tendon sheath — different problems, different injections.
For rotator cuff problems, the largest trial found no long-term benefit from subacromial corticosteroid, and no difference between a full exercise programme and a single advice session.
Frozen shoulder is the exception, and the one presentation where injection has the strongest standing.
Accuracy is worse here than at the knee, which is why the target question and the guidance question are related.

A patient told they are getting “a shoulder injection” has been told almost nothing. Four different injections go by that name, they treat different problems, and the evidence behind them is not equally strong.

The four targets

A simplification of a clinical assessment, not a substitute for one. Several of these coexist, and shoulder pain frequently comes from the neck rather than the shoulder at all.
Target	What points to it	Where the evidence stands
Subacromial space	Pain reaching overhead or behind the back, painful arc, night pain lying on it. The most commonly injected target.	Weakest of the four for long-term benefit — see below.
Glenohumeral joint	Global stiffness with loss of passive motion, particularly external rotation. Frozen shoulder or joint arthritis.	Strongest standing, especially in adhesive capsulitis.
AC joint	Pain localised to the top of the shoulder, tender to press, worse reaching across the body.	Small target, limited trial evidence, often diagnostically useful.
Biceps tendon sheath	Pain at the front of the shoulder, tender over the groove, worse with lifting or resisted flexion.	Little trial evidence; the accuracy case for guidance is strongest here.

What did the largest rotator cuff trial find?

The GRASP trial recruited 708 patients across 20 NHS trusts and tested two things at once: progressive exercise against a single session of best-practice advice, and corticosteroid injection against no injection. Outcomes were measured over twelve months.

Both comparisons came back null. There was no evidence of a difference between injection and no injection, and none between a full exercise programme and one advice session. The authors' conclusion: subacromial corticosteroid injection provided no long-term benefit in patients with rotator cuff disorders.

What that does and does not mean

GRASP measured outcomes over twelve months. It is evidence against injection as a durable solution for rotator cuff pain. It is not evidence that nobody gets short-term relief, and plenty of people plainly do.

The honest reading is the same one that applies at the knee: an injection buys a window. If the window is used, it can be worth having. If it is expected to fix a rotator cuff problem on its own, the largest trial we have says it will not.

Worth knowing alongside it: a Cochrane review of rotator cuff surgery found it may provide little or no benefit over non-operative treatment. This is a problem where the aggressive options have not outperformed the conservative ones.

Why is frozen shoulder different?

Because the problem is inflammation and contracture of the joint capsule itself, which is exactly what an intra-articular steroid acts on — and because the target is unambiguous.

The tell is loss of passive motion. If someone else moves your arm for you and it still will not go, that is a capsular restriction rather than a painful arc or a weak cuff. External rotation is usually the first and worst affected.

It is also the presentation where the injection most clearly enables something else: a shoulder that has become less painful tolerates the stretching that actually restores motion. On its own it does less.

Why accuracy matters more here

Shoulder targets are smaller, deeper and less reliably palpable than a knee. As covered in full on the guidance page, accuracy at the shoulder is more variable than at the knee, and for the subacromial bursa there is insufficient pooled data to state a figure at all.

That connects the two questions. If a shoulder injection did nothing, the possibilities are that the target was wrong, that the target was right and the needle missed, or that the treatment simply does not work well for that problem. Those have different next steps, and they are worth separating before concluding that injections do not work for you.

What I check before injecting a shoulder

Passive versus active motion, first. A shoulder that moves fully when I move it is a different problem from one that does not, and that single comparison redirects more shoulder injections than anything else I do.

Then the neck. Cervical radiculopathy presents as shoulder pain often enough that a shoulder examination which is entirely normal, in someone with real pain, should prompt looking elsewhere before reaching for a needle.

That is clinical reasoning rather than a trial finding — but it is why “which target” is the question this page is built around.

This article is general education about injection targets in the shoulder and the evidence behind them. It is not medical advice, does not create a patient–provider relationship, and cannot account for an individual situation. Distinguishing between these presentations requires a physical examination. For a specific person, that is a conversation with their own clinician. See the editorial policy for how this content is written and reviewed.

Sources

Hopewell S, Keene DJ, Marian IR, et al. Progressive exercise compared with best practice advice, with or without corticosteroid injection, for the treatment of patients with rotator cuff disorders (GRASP): a multicentre, pragmatic, 2 × 2 factorial, randomised controlled trial. The Lancet 2021;398(10298):416–428. 708 participants, 20 NHS trusts. Retrieved via PubMed.
Karjalainen TV, Jain NB, Heikkinen J, et al. Surgery for rotator cuff tears. Cochrane Database of Systematic Reviews 2019;12:CD013502. Retrieved via PubMed.