How close can two tags be placed?
Far enough apart that a phone can sit over one and not the other. In practice that means at least the width of a tag between them, more where the tags are small, and a visible target on each so people aim rather than guess.
The working rule
Leave at least the width of one tag between two tags, measured edge to edge, and treat that as a floor rather than a target.
Small tags need proportionally more, because the separation that matters is set by the phone's field rather than by the tags. Two small tags placed one tag width apart can both sit inside the region a phone excites.
Where space allows, a hand's width between tags removes the question entirely and needs no testing.
Why the field decides, not the tags
A phone produces a field over a region around its own coil. Any tag inside that region is woken, whether or not it is the one the person aimed at.
So the requirement is not that the tags do not touch. It is that a person presenting a phone at one tag does not bring the other into the field at the same time.
That region is larger than the phone's coil and it varies by model, which is why a layout validated on one device deserves a check on another.
When two tags are woken together the exchange has to be resolved, and a phone reading a tag normally is not built to offer a choice. The visible result is an unreliable read rather than a menu.
Rows and panels
A row of tags on a shelf edge, a control panel, or a set of labelled points on a machine is the common case, and it is where spacing gets squeezed.
Give each tag a printed target and a label, so the gesture is deliberate. Aiming at a marked spot is much more accurate than aiming at a region.
Stagger the rows if a grid is unavoidable, so no two tags are adjacent both horizontally and vertically.
And consider whether the row needs several tags at all. One tag leading to a page listing the options works on any spacing and never misreads.
Checking a layout
Test the worst case, which is the middle of the row rather than the ends. A tag with neighbours on both sides is the one that misreads.
Test with a phone whose coil sits low on its body, since that changes which tag ends up in the field when someone aims at a target.
Test the wrong gesture deliberately: approach at an angle, from the side, and too fast. If the layout survives that, it will survive the public.
And verify that each tag returns its own content, not merely that something reads. A neighbour answering instead of the intended tag is exactly the failure this spacing exists to prevent, and it looks like success unless you check what came back.
Frequently asked questions
Can tags touch if they are side by side?
They can be adjacent without interfering the way stacked tags do, and a phone will often wake both. Spacing is about the reader's field rather than about contact between tags.
Does a larger tag need more spacing?
Less, proportionally, because it is easier to aim at and easier to sit over exclusively. Small tags in a row are the hardest case.
What if I have no room to space them?
Use one tag pointing at a page that lists the options. It is more reliable than a cramped row and it costs nothing extra.
Do the tags interfere even when nobody is reading?
Adjacent tags sitting idle do nothing to each other, since there is no field. The interaction only exists while a reader is present.