Define buying situations
Agree the real questions, market, competitors and decision contexts worth testing. The prompt set is part of the methodology, not an invisible vendor default.
AI search evidence · for agencies
Your AI search platform measures the client. We test the claim you have to stand behind.
Your agency may already use Profound, Peec, Otterly, Semrush or another AI search platform. Keep it. When the client asks whether ChatGPT, Gemini, Perplexity or Google AI actually recommends them, and what the numbers really justify saying, Lumen & Lever runs an independent evidence study.
This service is for agencies. It is not an SMB monitoring subscription.
Initial study, pilot priceA$2,500 fixed
Controlled re-measureA$1,250
Evidence retainedResponses, citations, conditions
OutputWhite-label ready
The gap
A visibility score is useful. It is not the same thing as an answer you can defend.
AI search platforms already do valuable work. They track prompts, mentions, competitors, citations and change over time. We are not building another dashboard to replace them.
The problem appears when a client turns the dashboard into a harder question. Does AI actually recommend us? Are we genuinely improving? Why can I not reproduce this result? Did the work we did change anything? Can I put this in the board report?
Generative answers vary. Prompt wording, model, time, context and sampling can change what appears. One observation is not a market fact, and a correlation is not proof of cause.
We do not manufacture certainty from unstable observations.
We test a defined claim repeatedly, retain the evidence, and state what the observations support, what they suggest, and what they do not establish.
The study
A controlled evidence layer behind the client answer.
01
02
Observe repeatedly
Run controlled observations across the agreed AI answer surfaces and repetitions rather than treating a single generated answer as representative.
03
Preserve the record
Retain prompt, response, citations, timing, conditions, brand mentions, competitors and other agreed evidence so the conclusion can be checked later.
04
Find recurring patterns
Compare recommendation frequency, competitor presence, cited sources and recurring evidence relationships across the observation set.
05
Separate fact from inference
Observed, associated, inferred and unproven are not collapsed into one score. The report says which is which.
06
Re-measure
Repeat the same protocol after intervention so the agency can show what changed, without pretending that temporal association alone proves causation.
Works beside your stack
Keep Profound. Keep Peec. Keep Semrush. This is a different job.
Your monitoring platform is useful for continuous visibility intelligence. It may already retain prompts, answers, citations and history. We do not claim otherwise.
Lumen & Lever is brought in when the agency needs an independent, fixed-scope examination of a specific client claim or buying situation. The output is designed to sit behind the advice the agency gives, not to become another daily dashboard your team has to operate.
If your existing platform already answers the question to the standard you and your client require, you do not need us.
- Independent
We do not sell the visibility platform whose numbers are in question.
- Fixed scope
One defined claim, agreed before work starts, not an open monitoring retainer.
- Retained
Prompt, response, citation and condition evidence kept so the conclusion can be revisited.
- Bounded
No causal claim where the evidence establishes only correlation.
The output
What we saw. What it supports. What remains unproven.
The report is written so an agency can explain the result without hiding behind a proprietary score.
Observed
Competitor A appeared more often than the client in the defined buying situations.
Frequency measured across the agreed observation set, not inferred from one answer.
Associated
Source Y repeatedly appeared alongside recommendations for Competitor A.
The relationship is present in the evidence and worth investigating.
Not established
That changing Source Y will cause recommendation frequency to improve.
The current evidence does not establish causation. An intervention and a controlled re-measure would be required.
-
What the study delivers
The defined buying situations and observation protocol. Brand and competitor recommendation observations. A response and citation evidence archive. Recurring source and evidence patterns. Observed findings separated from inferred ones. Explicit limits and unresolved questions. A white-label-ready client report.
-
What it does not deliver
A guarantee about what an AI system will say next month, a proprietary score in place of the evidence, or a causal claim the observations cannot carry.
Pilot pricing
Price the evidence, not the screenshots.
-
AI Search Evidence Study · A$2,500 fixed
One client brand, a defined set of buying situations, agreed competitors and AI surfaces, repeated observations, citation and source analysis, retained evidence, a white-label-ready report and an agency debrief. Final scope is confirmed before work starts.
-
Controlled re-measure · A$1,250
Repeat the agreed protocol after the agency or client has acted. Compare the observation sets and report what changed, while keeping causation claims inside what the evidence can support.
-
Portfolio work
Quoted separately. We are testing this service with agencies now, so these are pilot prices rather than a manufactured market rate.
Why Lumen & Lever
Evidence discipline applied to AI search.
Lumen & Lever builds traceable AI systems for work where the answer has to hold up. The same standard applies here: preserve the source material, separate observation from conclusion, report uncertainty and keep a record that can be revisited.
The internal measurement tooling is ours. The client buys the study and the evidence, not another software login.
The service is led by Lee Powell. More about who leads it.
- Traceable
Every reported observation resolves to the retained response that produced it.
- Reviewed
Named human review before a conclusion is put in front of a client.
- Honest
Uncertain answers are marked uncertain. Confidence is reported, not performed.
For agencies
Bring the client question your dashboard cannot answer cleanly.
Tell us the brand, the market and the claim the client wants you to stand behind. We will tell you whether a controlled evidence study can answer it, and where it cannot. Write to hello@lumenandlever.com or use the form.