Method
How BackTier Measures AI Visibility
One cycle, repeated: measure, diagnose, implement, re-measure. Individual capabilities sit underneath it rather than being sold as separate products.
The four steps
- 01
Baseline
Freeze a prompt set for your category and record how each platform answers it today: entity accuracy, retrieval coverage, citation presence, answer accuracy, and which competitors appear.
- 02
Diagnose
Establish which constraint is actually binding — identity confusion, insufficient evidence, inaccessible content, weak third-party corroboration, technical structure, or category positioning.
- 03
Implement
Fix the priority issues: entity architecture, structured data, answer-ready content, internal linking, source development, and legitimate external authority building.
- 04
Re-measure
Re-run the original tests on a documented schedule and report what improved, what regressed, what stayed the same, and what cannot be attributed with confidence.
How the measurement is kept honest
The prompt set is written down before anything changes
We agree a fixed list of the questions your buyers actually ask, in the wording they use. That list is frozen so later readings are comparable to the first one. Adding prompts mid-engagement to make a chart look better is how visibility reporting becomes fiction.
Platforms are named, and treated as different systems
ChatGPT, Google AI Overviews, Gemini, Claude, and Perplexity retrieve and select differently. A change that moves one may not move another, and we report them separately instead of averaging them into a single score.
Readings are repeated, because the systems are non-deterministic
The same prompt can return different sources on consecutive runs. We record ranges across repeated runs and look at trends over time. A single citation is not evidence of a durable change.
Work is dated so it can be tied to the record
Entity architecture, structured data, retrievable content, and citation pathways ship in tracked increments with dates attached, so a movement can be examined against what was actually shipped rather than credited retroactively.
Reports say what did not change
Every report states what was measured, over what window, against which baseline, what improved, what regressed, what stayed flat, and where a movement cannot be attributed to a specific change.
What we will not claim
- No provider can guarantee what an independent AI platform will cite or recommend.
- Model and index updates happen outside anyone's control and can move results in either direction.
- Results vary by model, category, competition, and retrieval environment.
- Traffic and revenue effects depend on factors beyond AI visibility, so we do not present them as directly attributable unless your own analytics support it.