| UI or UX | UI + UX - Lifecycle notice for AI model version changes, retirement, migration, capability shifts, or replacement behavior | UI + UX - Page-flow service notification before the H1 | UI + UX - Severe-consequence warning copy before an action | UI + UX - Service phase and feedback marker | UI + UX - Calibrated reliability and uncertainty display for AI or automated predictions before user action | UI + UX - Whole-answer source coverage and grounding evidence display |
| UI guidance | Render model update notice as a lifecycle-specific communication that names the current model, new or replacement model, affected product surface, effective date, retirement or shutdown date, expected behavior change, migration path, test window, and whether action is required. | Render one notification banner in the page content column immediately before the page H1, using the same width as the surrounding service content. | Render warning text as a short high-emphasis statement with a warning icon, visible or hidden warning label, and explicit consequence copy placed before the relevant action, declaration, or instruction. | Render a compact phase tag such as Alpha or Beta with a short sentence and feedback link in the service header area, directly after service navigation or the primary header. | Render confidence and uncertainty as labelled reliability information with confidence band, reason, input scope, calibration status, review threshold, freshness, and the next safe action. | Render source grounding as an answer-wide evidence panel that separates source scope, searched sources, retrieved sources, used sources, supported claims, partially supported claims, unsupported claims, and unresolved source states. |
| UX guidance | Use model update notice when a model change can affect availability, output quality, latency, pricing, safety behavior, compliance, prompt compatibility, tool use, evaluations, or user trust. | Reserve the pattern for service-wide problems, personal deadlines or obligations elsewhere in the service, and outcomes from a previous page when the journey continues rather than ends. | Use warning text when users must understand a serious consequence before acting or failing to act, such as a fine, loss of access, permanent deletion, eligibility impact, or legal responsibility. | Use a phase banner to set expectations that a service is still being worked on and to invite feedback that helps the team improve it. | Use confidence / uncertainty display when users need to decide whether an AI prediction, classification, recommendation, extraction, risk assessment, or generated answer is reliable enough to apply. | Let users inspect coverage gaps, source scope, retrieval status, and unsupported claims before they copy, publish, apply, or rely on the answer. |
| Good UI | An admin console banner says GPT-4o 2024-08-06 retires on 2026-10-01, shows replacement GPT-5.1, lists three affected assistants, links to migration tests, and marks provisioned deployments as manual migration required. | A case-management page shows an Important notification banner immediately before Manage evidence, saying evidence upload may be delayed and linking to service status. | Before Submit declaration, a warning with an exclamation icon says the user may be fined if they provide false information. | A private beta benefits service shows a Beta tag directly below service navigation with Help us improve this service and a feedback link that preserves the current route. | A claim classifier says Medium confidence, 71 to 78 percent calibrated range, review threshold 80 percent, conflicting account-age signal, and routes the case to manual review. | A policy answer includes a Grounding panel showing 4 sources searched, 3 retrieved, 2 used, 5 supported claims, 1 partially supported claim, and 1 unsupported claim with a Review action. |
| Bad UI | A chat product silently changes model behavior overnight and users only discover the update after saved prompts produce different outputs. | A success toast appears in the corner after redirect and disappears before the user can verify which evidence was uploaded. | A red sentence says Important below the submit button after the user has already acted. | A bright sticky Beta pill floats over the form submit button and competes with validation. | Users over-trust an AI answer because a confidence badge looks authoritative. | The answer shows a green Grounded badge even though only one citation supports one paragraph. |
| Good UX | A developer sees a model retirement notice, opens the affected endpoint list, runs a saved regression set against the replacement model, schedules rollout, and dismisses the notice only after migration is complete. | A user lands on Evidence overview and sees the upload confirmation before the page heading, then can continue the unfinished service journey. | Users see the fine or eligibility consequence before checking the declaration and can pause to verify their answer. | A user in public beta sees the service is still improving, reports a problem from the current page, and returns to the same application step. | A reviewer sees low confidence and out-of-distribution input, opens the reason panel, collects the missing invoice, and avoids auto-denying the claim. | The person reaches the intended result with clear feedback. |
| Bad UX | A model deprecation email links to a changelog but the product UI still shows the retired model as healthy. | The same upload confirmation remains on unrelated pages after the user has already reviewed the evidence record. | Users incur a penalty because the warning looked like ordinary hint text. | A beta banner is hidden after the first visit, so deep-linked users miss the service maturity signal. | A model update invalidates old confidence scores but the UI keeps showing them as current. | Grounded status is shown before retrieval finishes. |
| Best fit | A model version, provider, lifecycle stage, availability, behavior, or replacement plan changes in a way users may need to understand or act on. | A service problem, deadline, overdue action, or elsewhere-in-service event affects the user's next interpretation of the page. | A user must understand a serious consequence before taking or skipping an action. | A service is in alpha, private beta, public beta, or another explicit pre-live phase. | Users must judge whether an AI prediction, classification, recommendation, extraction, risk score, or generated answer is reliable enough to use. | Users need answer-wide evidence coverage before trusting generated content. |
| Avoid when | The user only needs confidence for one answer or source evidence for one claim. | The message is a submitted-form validation error or field correction. | The message is a dynamic task status that must be announced when it appears. | Only one feature, experiment, or setting is in beta inside an otherwise live service. | Users only need source coverage or claim evidence, which belongs to source grounding display or citation display. | The only task is opening one claim's source preview, which belongs to citation display. |
| Required state | Deprecated model state with replacement model, retirement date, affected usage, recommended migration path, and owner. | Neutral Important state for service problem, deadline, overdue action, or elsewhere-in-service information. | Before-action warning state placed immediately before the related control or declaration. | Alpha state for limited prototype or research-only service. | High confidence state with calibration scope, reason, and whether direct apply is allowed. | Default grounded state with source scope, searched sources, retrieved sources, used sources, and supported-claim count. |
| Accessibility burden | Use accessible tables or lists for affected workflows, deployments, owners, and replacement test results. | Use a labelled region for neutral page-load notification banners so screen-reader users can navigate to the notice. | Keep the warning in reading order before the control, declaration, or instruction it qualifies. | Place the phase banner in reading order after the header or service navigation and before page-specific content. | Expose confidence label, uncertainty reason, threshold, freshness, and gated action as text rather than relying on color, position, or animation alone. | Expose grounding summary, source scope, status counts, unsupported claims, and source groups as text. |
| Common misuse | Using a dismissible banner that hides required migration before it is complete. | Using a notification banner for field validation errors instead of an error summary and field-level messages. | Using warning text for routine hints, explanations, or mild reminders. | Using Beta as a decorative badge for a feature rather than the service phase. | Showing a fake percent or exact decimal for an uncalibrated model score. | Showing a global Grounded badge when only some claims have evidence. |