Illustrative example
Every number on this page is made up, for a company called Northwind Signalsthat does not exist. It shows the shape of a real baseline, not a real customer’s results.
In one paragraph
A ModernTechLap baseline asks your tracked questions on every AI system you connect, stores each answer with the engine, model and date, and reports your AI Visibility Score, share of voice, recommendation rate, position and cited sources from that stored text. It then lists the actions worth doing first, each one pointing at the answers it came from. This example covers 25 questions on all 7 systems.
AI Visibility Score
21
out of 100, first run
Share of voice
24%
of all tracked-name mentions
Recommendation rate
12%
21 of 172 answers
Average position
3.4
when named at all
100 × (0.5 × 0.238 mention rate + 0.3 × 0.122 recommendation rate + 0.2 × 0.29 position score) = 21
41 of 172 stored answers named this brand, 21 recommended it, and the answers that named it averaged position 3.4. Every input is a count over answers you can open and read. The formulas are on the methodology page.
Fictional figures for a fictional company. One run is one sample: a real account repeats this on its plan’s schedule, and the dated series is the measurement.
One score across a whole category hides the thing worth knowing. Grouped by what the buyer was asking, next to the competitor doing best on the same questions.
| Question group | Questions | You | Best rival |
|---|---|---|---|
| Best-of: OT security platformsbehind | 6 | 18% | 71%Competitor A |
| Comparison: you versus a named rivalcompetitive | 4 | 64% | 68%Competitor A |
| Alternatives to the category leaderabsent | 4 | 0% | 54%Competitor B |
| Problem and solution: securing a plant networkcompetitive | 7 | 29% | 33%Competitor C |
| Purchase intent: pricing and procurementbehind | 4 | 9% | 60%Competitor A |
Each stored answer sits beside the question that produced it, the system that answered and the verdict read off the text. Five of 172 are shown here.
| Question | System | Verdict |
|---|---|---|
| Best-ofWhat are the best OT security platforms for manufacturers? | Gemini | MentionedNamed third of five. Position 3, recommended. |
| RecommendationWhich companies would you recommend for industrial IoT security? | ChatGPT | OpportunityTwo competitors named, you were not. |
| ComparisonHow does Northwind Signals compare with the market leader? | Perplexity | MentionedPosition 2, but with a negative qualifier, so not counted as recommended. |
| AlternativesWhat are the alternatives to the category leader? | Claude | OpportunityFour alternatives named, none of them you. |
| Purchase intentI need OT security for a mid-size plant. Which vendors should I evaluate? | Microsoft Copilot | OpportunityAnswer leaned entirely on one analyst roundup. No citation of yours. |
Share of voice is each name’s share of every mention of a tracked name, so these add up to 100%. The useful finding is rarely that you are invisible; it is which specific competitor owns the questions you should own.
Where a system returned its citations, they are recorded per domain. This is usually the most actionable section: it names the specific pages that decide who gets recommended. An answer that cited nothing is marked uncited rather than having sources inferred for it.
Each action names the answers it came from and carries an impact score, so you can disagree with the ranking and reorder it. On the Free plan you see these three and can read them; a paid plan gives you the whole backlog to edit, assign and mark done.
Why it matters. These are the questions a buyer asks while building a shortlist. Appearing on comparison questions but not alternatives ones means you get evaluated only once someone already knows your name.
What to do. Publish a sourced alternatives page naming the four vendors these answers name, with the trade-offs stated plainly and who each one suits.
Why it matters. Two systems described your category correctly. One called you a general network-security vendor, which puts you in a different shortlist entirely.
What to do. Make the one-line description identical on your site, your structured data and the third-party profiles these answers cite.
Why it matters. It was cited in 44 of the stored answers and does not list you. Whoever that page includes gets named; whoever it omits does not.
What to do. Get current, accurate information in front of that publication, and comparable coverage onto the two other domains these answers already read.
Mark an action done and the questions it targets are asked again about a week later. Both samples are dated and shown side by side.
Alternatives page published, 22 Aug 2026. Re-measured 29 Aug 2026.
Before
0%
4 alternatives questions, 14 Aug
After
18%
same 4 questions, 29 Aug
This is a comparison of two dated samples, not proof of cause. Model updates, competitor moves and ordinary run-to-run variance all sit inside that week, and where a change cannot be attributed the product says so instead of claiming the credit.
Free forever: 25 tracked questions on all 7 AI systems, a monthly refresh and your top three fixes. No card. ModernTechLap measures and influences the evidence around a business; it does not control AI model outputs, so nothing here is a promise that a system will name you.