Apple to add optional "Suggested Genmoji" keyboard toggle in iOS 27 using photos and typing history

Apple is testing an optional iOS 27 keyboard feature that will surface “Suggested Genmoji” generated from a user’s photo library and commonly typed phrases, according to a May 17, 2026 report in Mark Gurman’s Power On newsletter. The change puts Genmoji suggestions directly into the keyboard experience rather than relying solely on user prompts, which could alter how users discover and use the feature if it ships widely.

Genmoji first arrived in iOS 18.2 as part of Apple Intelligence, using image‑generation models to create emoji‑like images from typed prompts. Early iterations offered straightforward prompt‑to‑image creation but produced inconsistent results at times. Apple expanded Genmoji in iOS 26 with deeper customization and a mixing mechanic that lets users combine two emoji, adding more creative controls for users who manually initiate the feature.

For iOS 27 and iPadOS 27, the report says Apple exposed a keyboard‑level toggle labeled precisely, “Suggested Genmoji are created from your photos and your commonly typed phrases.” That setting is described as optional and located in keyboard settings; the intent appears to be boosting Genmoji adoption by offering predictive, context‑aware suggestions instead of waiting for users to type explicit prompts.

The proposed approach presents both opportunity and apprehension. Automatically generated suggestions could make Genmoji more useful if the models reliably produce contextually relevant results, but there is clear user concern about a feature that leverages intimate data sources — personal photos and typing history. Apple’s inclusion of a simple on/off toggle is presented in the report as the primary user control to mitigate those privacy worries.

The report does not confirm whether Genmoji’s image models will continue to run entirely on‑device in iOS 27. It also notes there was no explicit mention of an image‑model upgrade in the iOS 27 descriptions provided, which leaves on‑device inference a plausible default. That implementation detail matters for developers and users alike because on‑device models affect latency, compute constraints and the privacy surface compared with cloud‑based image inference.

Key uncertainties remain: the coverage ties the toggle to the broader iOS 27/iPadOS 27 software update but gives no specific release date or deeper technical specs. If Apple ships Suggested Genmoji as described, its impact will hinge on model quality, contextual relevance and how Apple balances privacy and compute trade‑offs when turning predictive suggestions on for users.