Pitch an Augmented-Reality Commercial Through What the Viewer Can Actually Do
Pitch an Augmented-Reality Commercial Through What the Viewer Can Actually Do
A lamp glowing above a kitchen counter is an afternoon in an image editor and almost no information about a commercial. The counter isn't a counter. The phone isn't a phone. Nobody has stood anywhere and placed anything. What the image sells is ambition, and ambition is cheap to draw.
It isn't wrong, exactly. It's a picture of the outcome with the entire event removed. And the event is what the client is actually being asked to buy into, which is why the second question in the room is always: and what does the person do?
A usable AR treatment answers six things. Where the person is standing and what they're holding. What they do first. What the experience reads from the room and what it merely draws on the screen. One complete interaction, carried through to something the viewer can see. A condition that defeats the idea, plus a way out that isn't a loop. And which parts of the demonstration actually ran. That last one is where pitches go soft, usually without meaning to.
Begin before the virtual object appears
The example running through this piece is invented: a furniture brand wants to advertise a table lamp with an AR experience. Somebody at home points a phone at a desk, places a lamp that isn't there, looks at it, and leaves. Nothing about it has been built. That's fine. A pitch is supposed to describe something unbuilt. What it can't do is describe the unbuilt thing as though it already works.
So open on a body. A person at home, at a desk, phone in one hand. No lamp in frame yet. Where did they come from? A QR code on the box flap is one answer. A link in a feed is another. They imply different rooms and different postures: the box scanner is already standing at the desk, real lamp still in its packaging, in a mood to assemble something. The feed tapper is on a sofa, mildly curious, and has to walk somewhere. Pick one, because that choice is the first frame of the ad and the first sentence of the treatment. This example takes the box flap: it puts the person at the desk already, with the real lamp within reach to compare against, which is the posture everything below assumes.
Then the camera permission dialog, which is the least glamorous frame in the deck and the one that decides whether anything after it exists. Say what happens to someone who taps don't allow. If the answer is "the experience ends," write that. That's a real audience with a real reaction, and a treatment that ignores them is a treatment that hasn't thought past the demo.
Then the surface, in plain language. The experience needs a clear, roughly flat area at least as wide as the lamp's base and deep enough that the lamp doesn't hang off the edge. Light helps. Clutter hurts. Describe the friendly version, and be honest that it's the friendly version.
One more physical fact that pitches forget. The phone occupies a hand. That hand is aiming and tapping, so the other hand is doing whatever else the idea requires — holding a mug, holding the real lamp for comparison, steadying a toddler. If your interaction needs a second hand, say so, because you've just committed the idea to a person who may not have one free.
The first action is small and specific: aim the phone at the desk until the experience signals it has found a surface. That signal is the earliest moment the commercial can say anything. Everything before it is paperwork.
Show what belongs to the room and what is added
Here's the distinction the whole treatment rests on. Some of what the viewer sees comes from the room. Some comes from the advertiser. Keep them in separate sentences and the rest gets easier to write.
Google's ARCore developer documentation describes motion tracking, environmental understanding, light estimation, and placing virtual objects in a physical scene as the working parts of an AR experience — the platform describing its own general model, as of the overview checked in September 2026. That tells you how tracking and environment understanding relate to each other. It tells you nothing about this lamp, this desk, or the phone in a particular viewer's pocket. It is also not a statement that every device supports the same things.
With that framing, the lamp example reduces to something a client can argue with. The desk's height, its tilt, and the clear area on top of it belong to the room. The lamp belongs to the advertiser. Light splits between the two columns, and the split is worth stating out loud: the room's light — how bright, coming from which side — is something the experience reads, which is the light estimation in that list. The lamp's glow is something the advertiser draws. The drawn glow doesn't land on the real desk. If the promise is a lamp that lights a kitchen counter, that part of the promise is a picture, not a behavior. The experience finds a surface and treats it as the floor for an object that only exists on the screen.
Why spend a paragraph on this? Because it sets the boundary of what the commercial can promise. You can promise that the lamp will be drawn, and where. You cannot promise it will sit convincingly on a glass-topped desk or a cloth with a busy print, because that depends on a room you don't control. Whether a glass desk actually defeats the placement is a thing to test, not a thing to assert — but the treatment should say the question exists.
And the composited phone mockup — the one with the lovely lamp and the real-looking kitchen — has not detected or measured anything. It's a drawing with a phone frame around it. Two words fix the whole problem: simulated view. Label it, and the rest of your deck becomes trustworthy, because you've shown you know the difference.
Follow one complete interaction
Take the interaction all the way through, once, in order. Short beats are better than a feature list.
The person aims the phone at the desk. The experience signals that it has found a surface. The lamp appears where they tap, sitting on the desk at a proposed size. They step to the left to see the shade from the side, and the phone moves with them, so the viewpoint changes: which side of the shade they're looking at, whether the lamp overlaps the mug already on the desk, whether the base looks planted or floating. What doesn't change is the lamp's position. It stays where it was put. It doesn't follow the phone, and it doesn't drift. That fixedness is the whole illusion, and it's worth one sentence in the treatment, because it's the thing the viewer will notice if it fails.
Then one meaningful action. In this invented example, the person taps the lamp to cycle three finishes: brass, matte black, white. The consequence is visible immediately — the object changes in place, same position, same apparent size. That's an action the commercial's idea actually needs, because the idea is which lamp belongs in your room, and finish is the decision a buyer is trying to make.
That's the test for every other gesture you're tempted to add. Pinch to resize, double-tap to spin, drag to reposition: each one costs the viewer attention and the pitch another paragraph of explanation. Include a gesture only when the commercial's idea requires it. A treatment with one well-explained action reads as confident. A treatment with five reads as nervous.
Note where a claim sneaks in. "The lamp appears at a realistic size" is a claim about scale, and scale relative to the real mug on the real desk is exactly the kind of thing a simulated view cannot show you. If nobody has measured it, the board says scale not verified, and the treatment stays honest without getting weaker.
The sequence ends with the person done: they close the experience with a control that stayed visible the whole time. Not a hidden gesture. Not a swipe from a corner. Something a person can find while holding a phone in one hand over a desk.
Include a condition that breaks the ideal sequence
Every AR pitch should carry one failure the viewer can recognize from their own home, not a technical edge case. Here, it's the clutter: the desk is covered in mail, or the only clear patch is narrower than the lamp's base. The friendly version of the surface requirement just met its opposite.
Choose a failure that matters to the idea, then write the proposed response and the exit. In this example, the response could be this: after a second failed placement, the experience stops asking the person to rescan the desk and offers to place the lamp at a smaller scale, with one line of plain explanation — the clear area is smaller than the lamp's base. That's a design intention, not a tested behavior, and the treatment should say so in those words.
The exit is the part that gets skipped. Three rules are worth writing down. The close control never disappears, including during a failure. A failure never sends the person back through the permission step or the entry point; the person is already here, and repeating setup is how an ad loses someone. And there's a fallback: show the lamp as a plain image on the screen, labeled as a plain image, so a person on a bad surface still gets an answer to the question they came with — even if it's a smaller answer than the one the ad wanted to give.
Then leave the rest of the failure question open, honestly. How often placement fails in real homes is unknown and depends on the homes. Whether a specific device route supports the interaction is a separate question with a separate answer. A treatment that claims a universal failure rate, or that every phone behaves the same way, has moved from pitching to guessing.
Choose a demonstration that matches the claim
Two artifacts, doing two different jobs. A storyboard or a simulation shows the proposed sequence. A recorded device test shows behavior that was actually observed. Nothing in the lamp example was run, so nothing in it gets to look like a recording.
The strongest single page here is an interaction board with three rows and no glamour whatsoever.
Sensed space. What the device reports about the room, drawn as a diagram: a rectangle for the desk plane, a marked region for the clear area the surface search would need. A diagram, not a photograph — which keeps private surroundings out of a document that will travel to people the viewer never agreed to.
Proposed graphics. The lamp as it would be drawn, in each of the three finishes. Labeled as a render, because that's what it is.
Untested scale. The lamp's dimensions against a real object, marked unverified until somebody measures it against the room conditions in the brief.
Put that board next to the beautiful floating lamp and the difference is stark in the useful direction. The floaty render says one thing: this could be pretty. The board says three: this is what we're reading from the room, this is what we're drawing, this is what we haven't measured. Clients will otherwise assume all three are the first one, and that assumption is what turns into an argument three weeks into production.
If the idea would capture the room or keep frames of it, the treatment says so in a line, plainly, because that's a client decision rather than a production detail. Keep it to a line. This isn't the place for an implementation lecture.
What remains to test is worth naming one by one: whether the finish swap reads as a change of product rather than a change of picture at arm's length; whether the lamp holds its apparent size as the person walks around it; whether a hand passing over the desk interrupts the placement. Each of those needs a qualified test on the device route the client cares about, in a room resembling the brief, before any of it appears as a claim.
End on the question you can't answer yet
Finish with the sequence in one breath: arrive from the box flap, allow the camera, aim until the surface is found, place the lamp on the desk, step around it, tap to change the finish, close. A person, a room, an action, a way out.
Then the labeled plan: storyboard for what's proposed, recorded device test only for what was observed, three rows distinguishing sensed space, drawn graphics, and unmeasured scale.
Then the last line, which should be a question rather than a promise — whether the device route this client cares about supports this interaction in the room the ad assumes. That sentence leaves the client with something to decide. A floating lamp leaves them with something to admire, and an image to forget.
Frequently asked questions
What six things should a usable AR treatment answer?
Where the person is standing and what they are holding; what they do first; what the experience reads from the room versus merely draws on screen; one complete interaction carried through to something visible; a condition that defeats the idea plus a way out that is not a loop; and which parts of the demonstration actually ran.
Why begin before the virtual object appears?
The entry point—QR code on a box flap versus a link in a feed—implies different rooms and postures, and that choice is the first frame of the ad. The camera permission dialog also decides whether anything after it exists, so the treatment should say what happens to someone who does not allow it.
What can be promised about an AR lamp and what cannot?
You can promise the lamp will be drawn, and where. You cannot promise it will sit convincingly on a glass-topped desk or a busy cloth print, because that depends on a room you do not control. The drawn glow does not land on the real desk; whether glass defeats placement is a thing to test, not assert.
How should simulations and device tests be labeled?
A storyboard or simulation shows the proposed sequence. A recorded device test shows behavior actually observed. A composited phone mockup has not detected or measured anything and should be labeled 'simulated view.' In the lamp example nothing was run, so nothing gets to look like a recording.
What goes on an interaction board, and why is it useful next to a beautiful render?
Three rows: sensed space, drawn as a diagram of the desk plane and clear area; proposed graphics, with the lamp render in each finish labeled as a render; and untested scale, with dimensions against a real object marked unverified. The floaty render says it could be pretty; the board says what is read from the room, what is drawn, and what has not been measured.