Skip to content
Interface language: English
Design / Signal

The site asks for the camera. Does the user know why?

How do you explain access to a camera, microphone or location? Design the moment of the request, waiting, refusal and another way to complete the task.

Article language: English

A woman beside a cup and folded red fabric by the water.
In this article

“This site wants to use your camera.” The customer only wanted to read how to report a damaged lamp. They have not found the form yet, but are already being asked to show part of their living room. The lamp is there, along with a pile of laundry and a person who never agreed to appear in a user manual.

The browser's request appears above the page, but its meaning is established earlier in the design. People should understand what action they started, why it needs a particular device resource and what they can do instead. “We need permissions to function properly” answers none of those questions.

Imagine a fictional lamp repair service. A customer can attach a picture of the model plate, describe the damage and choose a service location. The team is also considering a short voice note. Each feature uses a different device resource. There is no reason for all of them to trigger permission requests as soon as someone enters the page.

Show the action and its limits before asking

Beside the model number, offer “Take a photo of the plate”. Keep “Choose an existing photo” and a field for typing the model number, if the process accepts it. A short explanation says why the photograph is needed: it helps the service team identify the lamp. It does not imply that every number is read automatically if the service has no verified capability to do so.

web.dev's permissions guidance recommends asking after a user action, in a context people understand, and offering an alternative where possible. This is design advice, not a guarantee of a particular business outcome.

In our example, the customer first chooses the camera. The site's own explanation prepares them for the browser prompt without impersonating it. A “Continue to camera” button does not need to be called “I agree”: it does not replace the decision in the device prompt. Still less should it suggest acceptance of terms or every later use of the material. We are designing access to a feature here, not the legal conditions for handling data.

Show when sending occurs too. Starting the camera, taking a picture and submitting it with the case are separate stages. The customer may want to adjust the frame or abandon the picture after seeing a preview. If material is transmitted earlier, both the description and the behaviour must make that clear. Do not retain a “preview” label if it actually means the image is already attached to the report automatically.

Request the resource needed for the chosen feature. In our example, a photo of the plate does not need a microphone. If the customer chooses a voice note, explain recording separately. Bundling everything into the first request makes both the purpose and later problems harder to recognise: someone may deny audio when all they wanted was a photograph.

Labels on the page should match what happens next. After “Take a photo”, people expect the camera or an explanation of how to start it. “Add a photo” may open a file picker. If the design relies on different system behaviour, test it on supported devices. A drawing of a browser prompt in a mockup does not determine how a customer's phone will behave.

Well-prepared brand messages in an interface (in Polish) can be calm without hiding the effect. “Let's show your lamp in its best light” might work as a slogan on the page. Beside camera access, it is better to say what will start and why. A short description should warn that the camera will open; a joke can only supplement that explanation.

Denial is only one possible response

A design often has two views: the camera works, or the customer “did not give permission”. In a real task there are more reasons for a missing image. A suitable device may be absent, configuration may restrict access, or the user may never answer the prompt. The form should not wait forever saying “Starting the camera”.

MDN's getUserMedia documentation describes different causes of errors and the possibility that a prompt is left unanswered. A missing image does not always mean the user consciously refused.

Design materials need distinct situations, named for what is actually known:

  • Waiting for a choice. The page explains where the browser prompt may be and offers a way to another method. It announces neither a refusal nor a running camera.
  • Access blocked. The customer can change a setting or complete the task another way. The message must not promise that “Try again” will always open a new prompt.
  • No working image. If a device problem is identified, the text addresses the device. If the cause is unknown, the page preserves that uncertainty instead of blaming the customer.
  • Preview ready. The user understands whether they are only viewing the image or something is already being saved. They have clear actions to take a picture and finish.
  • Returning after changing settings. The damage description remains in the form. The page recognises the new state or explains what to retry without restarting the whole report.

MDN lists the permission states granted, denied and prompt; the last means permission has not been explicitly granted. It does not promise that a prompt will appear: attempting to use the feature may still be denied.

These names help implementers decide what should happen, but do not have to appear to customers as technical terms. “Access is blocked” says more than an error code. If it is not known whether the restriction follows a user decision or another condition, do not add “You blocked the camera”. Even a very small rectangle can make people feel wrongly accused.

Aeneia conceptual photography

When settings need changing, show instructions for the correct configuration. Do not ask someone to find a padlock where their browser shows a different symbol. More important than a detailed graphic is the ability to return to their own form. A customer may close the instructions, change a setting and still need the lamp description they already entered.

A refusal should lead to a complete alternative route through the task. “Choose a photo” needs its own file intake, and a manually entered model number should reach a useful place in the report. Attachment states (in Polish) do not disappear because the camera was requested first. The chosen image may still need to be uploaded and checked.

Not every alternative produces the same material. At the repair service, a model number alone may suffice for an initial question, but assessing visible damage will need a photograph. Explain the difference before submission. A customer without a camera can begin the conversation and supply a picture later if the company actually works that way. If the material is needed immediately, show the agreed help route instead of pretending any text can replace the evidence the team needs.

It also helps to distinguish a photograph from material already accepted. A customer takes a picture, sees it in the form and presses “Replace”. The design should say whether the previous file remains with the case until the replacement succeeds. Camera permission does not explain what happens to the earlier photograph. One permission can support many shots, but the customer needs to recognise the one selected for sending.

There is also a trickier sequence: the customer misses the browser prompt, chooses an existing file and only then allows the camera. A late response should not replace the selected picture with a new preview without explanation. Associate each result with the action that caused it and define what happens after that action is abandoned. Test this exact sequence during acceptance. A view tested only by immediately pressing “Allow” misses the conflict between the two methods.

What changes with the microphone and with location

In our repair service, a photograph helps identify the model. A voice recording would replace a written description. Before adding a microphone button, the team should decide who will listen and how the recording reaches the case. The ability to record does not mean the service team can use it later. A text alternative is especially useful when a customer does not want to speak aloud or has no quiet place.

After recording is selected, separate signals are needed: whether sound is being collected, whether recording has stopped and which material will be sent. A ready microphone does not mean a person has begun speaking. Use labels that match actual events. After cancellation, check whether the page has stopped using the microphone. Closing a colourful window alone does not prove it.

Before sending a voice note, the customer should be able to recognise what they prepared. Listening back, deleting and recording again each need clear actions. If the company offers transcription, a written record of speech, instead of audio alone, it must separately determine which material the service team receives and what the customer can correct. In our example the team receives the recording. The customer listens before sending and can replace it if it captured mostly the nearby radio rather than a description of the lamp.

Location serves another purpose: finding a service point. But a customer reporting a lamp from work may want to drop it off near home. Even an accurate device location does not decide where the person intends to go. Alongside “Use my location”, retain a town field or a choice from a list.

MDN describes obtaining a location as a feature requiring permission. The call may also be restricted by the site's policy; acceptable result age and waiting time have separate settings.

The customer needs a recognisable starting place and the correct service point. If location retrieval is pending or fails, the list of locations need not disappear. Do not call it “no service nearby” when the only missing information is the position. The design of a service-point map and list (in Polish) helps separate the device location from a location that actually handles the case.

Also go through the task after previously granted access has been revoked. A person may change settings between visits, which a prototype retaining the old state will not show. On return, check the prompt text, the route without that resource and what happens to a report already prepared. An earlier decision is not a promise for all future visits.

Notification settings (in Polish) require a similar distinction: choosing what a customer wants to hear about is a different decision from technical access to a channel. One “Enable everything” button can hide both. The lamp service should first establish how the customer wants an answer about a specific case, rather than request every possible permission.

One recognisable scenario is enough to begin acceptance, but run it by several routes. The customer photographs the model, abandons the camera, returns with a file, changes a setting and completes the report. Also test offline so that a local preview does not masquerade as an already submitted photograph. The article about offline work (in Polish) explains related states. In every run, the service team should be able to identify what it actually received.

Run the test without relying on icons alone. Action names should explain the camera, file selection and closing the preview. After returning from the browser prompt, customers must find where to continue when using a keyboard or screen reader too. If the preview fills the phone screen, provide a recognisable way to finish. Leaving should not require a gesture that was neither explained nor tested in the environment being used.

You can discuss the feature's purpose, the information the customer needs and the possible scope of work with Aeneia Design. In our example, a good ending is an accepted report with the correct model number, photograph or other agreed material. When finished, the customer sees the accepted material with their case, not another request to start the camera.

Aeneia / Aeneia editorial

Editorial article. Source links are included in the text.

About Aeneia

Brands we have worked with

  • L’Oréal
  • Garnier
  • Netflix
  • Carrefour
  • Warner Bros. Discovery
  • Biedronka
  • HBO Max
  • Wolt
  • Żabka
  • COSRX
  • Embassy of China in Poland
  • Mixa
Aeneia / Initium

From knowledge to a decision

Tell us what you want to put in order. We will agree on where to start.

Contact Aeneia

We partner with and use solutions from:

  • Google
  • OpenAI
  • Microsoft
  • Stripe
  • Anthropic
  • Linear