Updated on 20 October 2026.
Changing AI monitoring providers should not look like an uninterrupted history of one metric. Preserve material you may transfer, compare configurations and run both services under documented conditions. The goal is acceptance of the new service, not finding the dashboard with the highest percentage.
Establish what can move
Before cancelling the old service, establish access to questions, configuration, responses, notes and history. Check export scope, contractual terms and permission to handle client data. An available file does not mean the destination can import it. Keeping an archive outside the platform may be the available option.
Record missing material and when access expires. Do not promise that all history is portable or that none of it can move. Formats, features and provider arrangements determine the answer.
Build a compatibility matrix
| Element | Compare | Acceptance condition |
|---|---|---|
| Questions | Wording, versions, groups, branded and unbranded questions | Approved mapping between old and new entries |
| Environment | System, mode, collection method, country and language | Differences explicitly labelled |
| Observation | Timestamp, timezone, cadence and missing-data treatment | Scope can be reconstructed |
| Recognition | Brand, aliases, competitors and rules | Representative answers checked |
| Metric | Unit, numerator, denominator and aggregation | Equivalence demonstrated or series separated |
| History | Available material and format | Recipient can open retained files |
Plan the parallel comparison
Before starting, define the decision: access to answers, markets, sources or team workflow. Fix the questions, shared window, owner and completion criteria. Choose scope for the task and available allowance. No universal question count or duration guarantees an answer.
Do not change question wording and provider simultaneously without recording both changes. Preserve methodological differences even when metric labels match. Two providers are not randomised treatment groups for a content experiment. Parallel monitoring alone does not establish an article's effect on visibility.
Decide when to switch
Choose the provider that meets written requirements, supplies needed evidence and supports daily work. A higher percentage alone does not demonstrate better measurement. Keep separate series where definitions differ; mechanical rescaling does not reconcile methods.
Close migration after configuration, access, archive and initial valid observations are confirmed. Mark the provider change as a boundary in the history. Do not cancel the old service before agreed acceptance or promise automatic transfer of results.
Starting with Semly
When moving to Semly, check supported systems, question country and language, and plan scope. Transferring configuration, adding questions and publishing articles are separate actions. Confirm import options and history access before migrating rather than assuming them.
This is a proposed migration acceptance protocol, not a completed provider test or client results.
