How to track AI ads

Choose the brand and product category, then record the platform, country, language, and time window you intend to compare. Use the coverage actually available in your GeckoCheck workspace. Record missing coverage rather than treating it as a result.

Build a stable prompt set around real customer decisions. Include category discovery, comparisons, and buying constraints. Keep branded prompts separate from unbranded prompts so you know whether the user already named your brand.

• Read the AI ads visibility summary, then inspect the competitor and prompt details available for that period.
• For each observation, record the exact prompt, date, platform, market, sponsored label, advertiser, promoted product, and visible placement. Retain supporting evidence when available.
• If a field is not supplied, mark it unknown. Do not reconstruct a supposedly observed placement from a count alone.
• Keep the organic answer, brand mentions, and source links separate from the sponsored unit.

Use separate outcomes for an ad observed, no ad observed in a successful observation, and observation unavailable or failed. A missing reading is not a confirmed absence. An ad-free account, unsupported region, or different context can affect what is observable.

A recorded prompt is not every private conversation on the platform. Ad personalization and eligibility can produce different outcomes for different people. Avoid describing the tracked sample as a census of all users.

Review the same prompt cohort and conditions in the next period. Note any added prompts, changed coverage, or unavailable observations alongside the comparison. Count repeated observations consistently and distinguish occurrences from unique creative variants if your evidence allows it.

Example workflow: compare the same carry-on buying questions in the same market before deciding whether a new competitor message is a recurring pattern. The example is a suggested method, not a scheduled refresh promise or measured finding.

• Investigate missing or changing results ›