Methodology

How the score is calculated

Every number in AIgnored comes from a question we asked an AI model and an answer we can show you. This page explains the whole pipeline, including where it is approximate. If something here would change how you use the product, it is here on purpose.

1. Where the business list comes from

We ask the Google Places API for businesses matching your niche in your city, using your own API key. We keep only businesses that Google marks as operational and that have at least 5 reviews, because a business with no reviews is not an established business worth pitching.

We store the Google place_id, which Google allows indefinitely. During the scan we also keep a working snapshot of each business: its name, rating, review count and website domain. Matching AI answers to businesses, grading and the outreach draft need values to work on, and the snapshot is deleted with the scan. For a saved lead still marked new, contacted or replied, the snapshot is refreshed with your Places key once it is more than 25 days old. When you open a results table, those fields are fetched live with your Places key and the live values replace the snapshot on screen. That is why the table has a short loading moment, and why a table opened without a Places key shows no names, ratings or websites. The CSV export does not use the snapshot either: it looks the Google fields up live when you click. The privacy policy lists exactly what is kept.

This is an official API. AIgnored does not scrape Google, does not run headless browsers, and does not use third-party scraper services.

2. The questions we ask

Each scan asks 8 questions per model, phrased the way a real person would ask them, for example "Who is the best plumber in Austin, TX?" and "I need a plumber in Austin, TX. Who should I call first?" Asking one question once would measure luck. Asking 8 different phrasings measures a pattern.

The 8 default questions, for plumbers in Austin, TX:

Who is the best plumber in Austin, TX?
I need a plumber in Austin, TX. Who should I call first?
What are the top rated plumbers in Austin, TX right now?
Recommend a reliable plumber in Austin, TX for an urgent job.
Which plumbers in Austin, TX do locals actually recommend?
Who are the most trusted plumbing companies in Austin, TX?
If I wanted the highest quality plumber in Austin, TX, who would you name?
List the best plumbing businesses serving Austin, TX.

These are the defaults. An admin can change them, and each scan's why drawer shows the exact questions it asked.

Web search grounding is switched on wherever the provider supports it, so answers reflect what the model can find today rather than what it memorised during training. Each answer ends with a line listing the business names it gave, which is how we read them reliably.

3. Matching names to businesses

AI answers give names as people write them, not as Google stores them. We normalise both sides by lowercasing, stripping punctuation and accents, and dropping filler words such as LLC, Inc, The and Company. Full-width characters, trademark and copyright signs, and letters such as ß and Ł are folded too, and names in other scripts are compared as written. Then we compare the remaining words with a Dice coefficient, which scores the overlap between two sets of words from 0 to 1.

A score of 0.85 or higher counts as a match. Between 0.70 and 0.85 the match is listed as a possible match and left out of the score until you confirm it. Below 0.70 we ignore it. One exception: if every word of a name with two or more words appears in the same order inside the other name, it is listed as a possible match at 0.70 even when the overlap score is lower. That catches Abacus Plumbing inside Abacus Plumbing, Air Conditioning & Electrical. It also flags pairs such as Austin Plumbing inside South Austin Plumbing Heating & Air, which is why a person confirms possible matches before they count. You can see and change every borderline call, and confirming one recalculates the score immediately.

4. The visibility score

For each business we count how many distinct AI answers mentioned it, then divide by the number of answers that completed, and round to a whole number. A business named in 3 of 8 answers scores 38. A business named in none scores 0.

Grades sort the list into who is worth contacting:

A: score of 0, rating 4.2 or better, at least 25 reviews, and a website. A real business that the AI never names.
B: score of 15 or below, rating 4.0 or better, at least 10 reviews.
C: score of 40 or below.
Already visible: everything else. These are the businesses the AI does recommend.

The top competitor shown on each row is the business with the highest score in that scan, which is the name the AI gives instead.

5. Limits, stated plainly

API answers are close to, not identical to, the consumer apps. We query the same models through their APIs with search enabled. The ChatGPT, Gemini and Perplexity apps add their own system prompts, personalisation and product logic that no API exposes. Treat a scan as a strong approximation of what those apps say, not a transcript of them.

Models are not deterministic. Ask the same question twice and the names can change. That is why we ask 8 times per model, and why a score of 25 versus 38 matters less than a score of 0 versus 50. Re-running a scan a week later will move numbers. Businesses at 0 tend to stay at 0.

Matching is not perfect. Two genuinely different businesses can share most of a name. Borderline calls are surfaced rather than guessed, which is what the possible matches panel is for.

The business list is only as good as Google. A business with no Google profile, or fewer than 5 reviews, will not appear in a scan at all.

6. What a scan costs you

Scans run on your own API keys, so you pay the provider directly and we never mark it up. A scan is 8 AI calls per model, so 8 to 24 in total, plus 1 to 3 Places searches. Opening a results table makes one Place Details call per business, up to 60, and opening the leads board makes one per lead it shows. A CSV export makes one per business in the file, plus one for each top competitor not already in it. Once a day, a saved lead still marked new, contacted or replied whose snapshot is older than 25 days is refreshed with one Place Details call on your key. On Gemini's free grounding allowance the AI calls cost $0. On a paid model, multiply 8 by that model's published price per call. We do not quote a figure here because provider prices change and we would rather you read theirs.

Read the AI keys guide and the Google Places guide for the exact setup.