ChatRank by SwiftAppLab

SwiftAppLab / ChatRank / How it works

Methodology

How ChatRank measures AI search visibility.

A transparent, reproducible score. No black box. Here's exactly what happens when you type your domain into ChatRank — from scraping your homepage to the final 0-100 number.

Run your free check →

The Score

The 4-part visibility score

Your ChatRank score is a weighted composite of four signals that together capture how well your brand actually performs when a real buyer asks ChatGPT a real question.

40%

Mentions rate

Of the five buyer-intent prompts we run, what percentage return your brand by name? This is the single strongest signal of AI visibility.

25%

Average position

When ChatGPT returns a list, are you in the top three or buried at position eight? Higher rank maps linearly to a higher score.

15%

Sentiment

We analyze the exact wording ChatGPT uses about your brand and normalize it from -1 to 1, then map it to a 0-1 contribution to your score.

20%

Share of voice

Across all five prompts, your brand's mentions as a percentage of all brand mentions. Measures dominance vs your competitive set.

Score = (0.40 × mentions_rate) + (0.25 × position_score) + (0.15 × sentiment_norm) + (0.20 × share_of_voice)
All inputs normalized to 0-1, final score scaled to 0-100.

The Pipeline

What happens when you run a check

Five deterministic steps. Every run is logged so you can audit the exact prompts, responses, and citations behind your score.

  1. 1

    Scrape your homepage

    We fetch your domain and extract the primary headline, meta description, and the first two H2s. This gives us enough signal to understand what you sell without needing an onboarding form.

  2. 2

    Generate 5 buyer-intent queries

    Using the scraped context, we synthesize five prompts a real buyer would type — for example, "best AI receptionist for small business" or "how much does an AI answering service cost". Queries are listicle-style because that's what drives ranking consequences.

  3. 3

    Query ChatGPT with web search enabled

    Each query is sent to ChatGPT with the web search tool active — the same configuration a prospect uses in the ChatGPT app. We capture the full response text, any citations, and the source URLs ChatGPT pulled.

  4. 4

    Analyze responses

    For each response we extract: whether your brand is named, the position in any list, the sentiment of the surrounding sentence, competitor brands mentioned, and the cited sources. All parsing is deterministic and logged.

  5. 5

    Score & render the report

    We apply the 40/25/15/20 formula, generate the public report page, and stream results back to your browser. Total wall-clock time for a fresh run is usually under 30 seconds.

Scope

Why ChatGPT-only, for now

We'd rather do one engine well than four engines badly.

ChatGPT currently represents the largest share of AI search traffic in North America and most of Europe. Most of our customers' prospects are asking ChatGPT first, which makes it the highest-leverage engine to measure.

Perplexity support is the next engine on our roadmap — its citation-forward design makes it a natural second. Google AI Overviews and Gemini will follow once their APIs stabilize enough to give a reproducible measurement. We'd rather ship one engine that you can trust than four engines that give you inconsistent numbers.

If you need multi-engine coverage today, we're honest about it: Profound and Otterly cover more engines, at higher price points.

Transparency

How we handle caching & cost

LLM calls are not free. Here's exactly how we keep the free tier free without degrading the signal.

24-hour result cache, deterministic prompts

If your domain was checked in the last 24 hours, we serve the cached result instead of re-running the pipeline. AI search rankings don't change by the minute, and this keeps the tool usable for everyone. If you need real-time re-runs, paid plans override the cache on a schedule that fits your tier.

Rate limit: one free check per IP per hour. We surface this up front rather than hiding it behind a "something went wrong" message. Paid plans have no rate limit.

Why this is fair: the free check costs us real money in API fees. Caching plus rate limits let us give you a legitimately useful result without the signup wall most tools use. Your paid subscription is what keeps the free tier alive for the next person.

Improve your score

What you can do to rank higher in ChatGPT

Generative engine optimization isn't magic. Three things consistently move the needle — and the paid plans let you track which ones are working for you.

1

Publish structured, factual content

ChatGPT cites pages that answer the question directly. Comparison tables, pricing pages with exact numbers, feature grids, and clean FAQ sections get quoted. Marketing fluff does not.

2

Earn mentions in trusted sources

Wikipedia, high-authority Reddit threads, G2, Capterra, TrustPilot, and industry-specific review sites are what ChatGPT reaches for when it browses the web. One placement in the right thread can flip your score 10 points.

3

Add FAQ schema & entity markup

Structured data gives AI engines a cleaner signal about what your pages contain. FAQPage, Product, and Organization schema are cheap to add and meaningfully improve citability.

Ready to see your score?

One field, thirty seconds, no signup. Your first check is always free.

Run your free check → See pricing