Explain in plain words: Apple Intelligence explanations in AllergenLens

A finding such as E322 (lecithin) can come from soy or eggs is precise, but it doesn't say what lecithin is or why a label might not name the source. AllergenLens can add a few sentences in plain words, written on the device by Apple Intelligence where it's available. This post explains when the button appears, what the explanation is based on, and what it will never do.

Where the button appears

Open any result and look at the findings. Under each finding in Contains, May contain or Couldn't read clearly, you'll see Explain in plain words with a sparkle icon. VoiceOver reads its hint: Written by Apple Intelligence on this iPhone.

There is no button on Not found in the text we read. Nothing was found, so there's nothing to explain, and an explanation there could read like reassurance. Not found is not a guarantee either way.

The button only shows when all of these are true:

On other iPhones the button simply isn't there. Every finding still has its reason, and every warning still shows.

What the explanation is based on

The model isn't asked to judge the product. It is given only the finding and what produced it:

• A word-list match. For most findings, the model gets the word read on the label and the allergen word list term it matched, such as whey for milk or E322 with the list's own note. The basis line reads like Based on “whey” in the allergen word list v2. AI can be wrong. Always read the label.
• The app's own notes. For an Unclear finding, or for a word from Your own ingredients that the word list knows nothing about, the model gets only the reasons the app already shows. The basis line reads Based only on the reasons above. AI can be wrong. Always read the label.

The explanation is short, at most about 60 words, and written to stay close to the reason. When one appears, the Sources for that finding open too, so you can see the public sources the allergen lists are based on. The full list is under Sources and citations in Settings.

What it will never do

AllergenLens checks every explanation before showing it. The text is thrown away if it:

• sounds like reassurance, for example promising the food is fine;
• says what the product does not contain;
• plays the finding down, such as calling it unlikely;
• is too long, or doesn't mention the words it was asked about.

In that case you see This explanation wasn't shown because it didn't follow the app's wording rules. The reasons above still apply. If Apple Intelligence can't write anything right now, you get a similar note that the reasons still apply.

Most of all, the explanation never changes the result. Contains, May contain and Unclear come from the word list and the label text, the same on every iPhone. More on how those are chosen in result states and coverage, and on the words behind them in E-numbers and unclear words.

FAQ

Download AllergenLens and get plain-words explanations on supported iPhones.