How we measure.
What do we do, and how? What can we not measure? We explain it all on this page.
We prepare the questions.
We write the questions your customers would really ask. We look at your industry, your city and your services. We also add the topics you wrote in Onboarding. The questions contain no personal data. We also ask 10 reference questions to learn about your business and its surroundings.
90 questions in each scan. Each plan tracks one business. This is your domain name, or your business name with its city and district. On Maxxed Tracking and Enterprise Max, the next scan can have up to 270 questions.
We ask each question several times.
We ask ChatGPT, Gemini, Claude and Perplexity. Each question is asked in a separate session, one after another. On Maxxed Tracking and Enterprise Max, we also measure Google AI Overviews and Copilot. doodlis systems do this automatically.
We take ChatGPT, Gemini and Copilot answers from the app screen that everyone sees. Data providers collect them. Perplexity answers come from the app screen or from the API. We get Claude answers from the official API, with web search. The method note of each result says which way we used.
We read and examine the answers.
Which businesses were recommended, and in what order? Which sources did the answer rely on? Did your name come up? Is there wrong information about you? We open the source pages shown in the answers with our own downloader. While doing so, we follow each site's robots.txt rules. If our downloader cannot open a page, we try a reading service, only as a backup. We never open a page that robots.txt blocks.
We use AI to help us examine the answers. We check the result before it goes live, then put it in your dashboard.
We make two comparisons.
We compare the answers where your name came up with other answers to the same questions, and you with the recommended businesses. Every gap comes with two values: yours and the value usually seen among the recommended.
10 reference questions decide who you are compared with and by which criteria. We do not add these questions to your result, and we do not charge for them.
We test that a gap is not chance.
We account for chance across many comparisons. A gap that fails these checks is not shown to you.
- Enough examplesA gap resting on few answers is not reported.
- Chance correctionMany comparisons produce chance gaps; we correct for that.
- ConsistencyThe gap must hold across models and question types.
- Other factors held constantWe test while holding factors such as model and question type constant.
- We test ourselvesShuffle the labels at random and no result should appear. If one does, the method is fixed.
We give every gap as two numbers and one sentence.
The recommended usually have 528 reviews, you have 61. Every new review narrows this gap.
Sample dataNext to the result, we explain how to read it. When the result is ready, we let you know by email.
Every Friday we look at the general trend.
We track which kinds of sources AI relies on, and how much. We also collect this data automatically.
How often your analysis is updated
Each plan tracks one business. The first scan starts once we receive your payment and Onboarding is complete. On tracking plans, your dashboard shows the time left until the next scan.
- Single Scan: once, 4 models
- Light Tracking: every 2 weeks, 4 models
- Normal Tracking: once a week, 4 models
- Maxxed Tracking: every 3 days, 6 models
- Enterprise Pro: once a week, 4 models
- Enterprise Max: every 3 days, 6 models
What you see in your dashboard
- Analysis results: two values, the gap between them and one sentence
- All collected answers, by date and model
- The recommended: businesses recommended for the same questions
- Recommendations, an ordered to-do list and the journey of each task
- On tracking plans, new results and the next step with every update
- On tracking plans, Position: your rank in three separate frames and the points left to the next rank up
On Maxxed Tracking and Enterprise Max plans, the result also comes with a more detailed view. What it includes is written on the pricing page.
How Position's three rankings are built.
Each ranking has a frame and an objective membership rule. The scoring formula is the same for everyone.
- Every business that fits the rule is listed
- At least five businesses in each ranking
- One is always the widest frame
- If a frame changes, the reason is written
What we promise.
We keep these in every plan and on every screen.
- Every number rests on real answers
- Collected answers stay in your dashboard
- If the method changes, we say so
- No fake reviews, no hidden ads
What we cannot promise.
AI answers are produced by other companies. So there are things we do not promise.
- A specific rank
- A specific visibility result
- When an answer will change
- The single cause of a gap
Our limits
- AI answers can change by person, place, time and model version. Our scan is the result of questions asked at a certain moment, under certain conditions. Not every user may see the same answer.
- AI companies change their models without telling us. A business that appears in one scan may not appear in the next.
- We infer the weight of sources from the sources shown in the answers and from the content of the answer. We do not know how the companies work inside.
- Google AI Overviews is the AI summary at the top of Google search. Google does not show it for every search. Questions with no overview are not part of this model's analysis.
- If a model cannot be measured in one scan, we say so in the result. We run the analysis on the answers of the other models measured that day.
- The analysis is not a judgment about the recommended businesses. A gap only shows what is seen together in the answers; it does not prove the cause.
- The models decide which business comes up in an answer. We measure, find the reason and prepare the to-dos.
The full explanation of the service scope and results is in the Scope of Service and Results Statement document.
What the numbers mean.
What every number on our site and in your dashboard stands for.
- Answers analyzed
- All answers to the questions we ask AI models that enter the analysis. Trial and test scans are not counted.
- Answers analyzed this week
- Answers that entered the analysis in the last seven days.
- Number of sectors
- Distinct sectors with at least one analysis.
- The recommended
- Businesses recommended in answers to the same questions.
- Usual among the recommended
- The middle value when the recommended businesses' values are sorted. Extreme values do not move it.
- Gap
- The distance between your value and the comparison value. Not a verdict, a range that can be closed.
- Basis
- How many answers and businesses a result rests on.
- Coverage
- How many questions a result concerns.
How we use data.
Sector results use aggregate numbers with no names, domains or text. No other business sees your data.
Sample question list
The first 10 questions of a 90-question sample set prepared for a dental clinic in Brooklyn.Sample data
- Which clinic should I go to for dental implants in New York?
- Can you recommend a good dentist in New York?
- Where should I go for teeth whitening in New York?
- Which clinic do you recommend for dental implants in New York?
- Which clinic do you recommend for zirconia crowns in New York?
- Which clinic do you recommend for clear aligners in New York?
- Which clinic do you recommend for a smile makeover in New York?
- Which clinic do you recommend for orthodontic treatment in New York?
- Which clinic do you recommend for porcelain veneers in New York?
- Which clinic do you recommend for a root canal in New York?
