AI RANKINGNEW

Your real GSC queries. Run through four AI models.

AI Ranking pulls the buyer queries you already win in Search Console, strips out anything containing your brand or a competitor's, and measures whether ChatGPT, Claude, Gemini, and Perplexity cite you for the survivors. Every result is labeled with one of four citation states. No synthetic prompts. No brand name in any query. No guesswork.

§ 01Why it exists

Most AI visibility tools test prompts. AI Ranking tests evidence.

Every other AI visibility tool starts the same way: a marketer types in a list of prompts they think buyers might ask, then watches whether AI mentions the brand. The prompts are guesses. The measurement is theatre.

AI Ranking starts from your Search Console. The queries you already get organic impressions on are the queries buyers are already asking. No guessing. No synthetic prompt curation. No drift between what you measure and what your buyers actually search for.

Then the 5-layer brand-safety classifier removes every query that contains your brand, a competitor's brand, or your domain stem. The first layer is a word-boundary regex. The fifth is an embedding cosine similarity check against a learned vector of your brand. Nothing slips through. By the time a query reaches an AI model, it is a clean buyer-intent question with no name in it.

The top 30 highest-impact survivors fan out to ChatGPT, Claude, Gemini, and Perplexity. Every result is labeled with one of four citation states. Cited with URL. Name only. Descriptive only. Not mentioned. That is the unbiased measurement of where you stand.

§ 02How it works

Four stages. From Search Console to citation map.

01
Pull Search Console queries

Connects to your Google Search Console (read-only) and pulls every query that drove an impression in the last 28 days. No prompts to write, no keywords to maintain — your evidence is your real organic traffic.

02
5-layer brand-safety filter

Removes every query containing your brand name, a competitor's name, or your domain. Five cascading layers — regex, Levenshtein, batched GPT-4o, pre-flight re-check, and embedding cosine similarity — guarantee zero brand-name leakage to the AI models.

03
Top-30 fanout to 4 models

The 30 highest-impact survivors fan out to ChatGPT, Claude, Gemini, and Perplexity. Each (query × model) leaf is logged with the full model response, latency, and token cost.

04
Label every result

Each response is classified into one of four citation states: cited_with_url, name_only, descriptive_only, or not_mentioned. The dashboard shows a per-query breakdown, a per-model breakdown, and the raw response text behind every label.

§ 03Capabilities

Six things competitors can't match.

Real GSC queries, not synthetic prompts

AI Ranking only ever measures queries you already get organic impressions on. The result is grounded in your actual search demand, not a marketer's guess about what buyers might ask.

Zero brand name in any query

The 5-layer classifier guarantees no query reaching an AI model contains your brand, your competitor, or your domain. This is the only honest way to measure visibility — typing your own brand into ChatGPT is like asking a mirror if you're handsome.

4-state citation taxonomy

Every result is labeled cited_with_url, name_only, descriptive_only, or not_mentioned. Descriptive_only mentions — where AI describes your category position without naming you — are the recoverable layer most other tools miss.

Top-30 impact-scored queries

Survivors of the classifier are ranked by impact (impressions × position-weighted CTR potential) and capped at 30 per run. You measure the queries that move the needle, not every long-tail tail that GSC happened to log.

Chained-tick architecture

Runs are split into short-lived state-machine ticks instead of one long process. The pipeline survives Vercel's 300-second function ceiling, recovers from transient failures, and surfaces real-time progress via 4 stage labels and a percent-complete indicator.

Cost-protected weekly cap

Starter 20 runs/week, Growth 50, Agency 100, plus a 7-day cooldown per brand between manual runs. Hard ceiling of 200 queries per run regardless of tier. Cost protection is built into the architecture, not bolted on.

§ 04Who it is for

Anyone with active organic search demand.

SEO professionals adding the AI overlay

Traditional rank tracking covers Google. AI Ranking covers what AI models say to the same buyer queries. Layer it on top of your existing GSC reporting to show clients which of their winning organic queries lose visibility inside ChatGPT and Perplexity.

SaaS marketers measuring AI share-of-voice

Your top 30 GSC queries already tell you what category buyers ask about. AI Ranking shows you which of those are now answered by AI — and whether your brand is in the answer or your competitor is.

Agencies running AI Ranking across clients

Each client's run uses their own GSC, their own brand-safety terms, and their own competitor list. Agency tier gives 100 runs per week and admin-force re-runs for one-off urgent reports.

Marketing leaders with active GSC

If you can see organic impressions in Search Console, AI Ranking can measure your AI visibility on those exact queries. No new keyword research, no manual prompt curation, no maintenance.

§ 05Pricing

Included on every paid plan.

Starter
$49/mo

20 AI Ranking runs per week, 4 AI models, 7-day cooldown per brand.

Growth
$149/mo

50 AI Ranking runs per week, 4 AI models, 7-day cooldown per brand.

Agency
$249/mo

100 AI Ranking runs per week, all 4 AI models, admin-force re-runs available.

See full pricing →
§ 06Common questions

AI Ranking questions.

Why use my real GSC queries instead of pre-built prompts?

Pre-built prompts are guesses. Real Search Console queries are evidence — these are the questions buyers are already asking that drive traffic to your site. AI Ranking measures whether AI models recommend you for the exact queries your buyers actually use, not the ones a marketer at TopSlot imagined.

How does the 5-layer brand-safety classifier work?

No query containing your brand name, a competitor's name, or your domain stem is allowed to reach an AI model. Five layers run cheapest-first: word-boundary regex, Levenshtein typo catch, batched GPT-4o classification, a pre-flight regex re-check inside the model client, and embedding cosine similarity against a learned brand vector. The first layer that fires excludes the query. The result: an unbiased measurement of whether AI cites you when buyers don't already know your name.

What are the four citation states?

Every (query, model) result is labeled as one of: cited_with_url (your domain appears as a link), name_only (your brand name appears verbatim with no URL), descriptive_only (you're described in a way that semantically matches your brand without naming you), or not_mentioned. The four states give you a much sharper picture than binary mentioned/not-mentioned — descriptive_only mentions are recoverable into named citations with the right schema and content fixes.

Why is there a 7-day cooldown between manual runs?

AI Ranking runs are expensive — a single Growth-tier run fans 30 queries across 4 models for 120 model calls and pulls fresh data from your Search Console connection. A 7-day cooldown per brand keeps your weekly cap meaningful and prevents accidental cost runaway. Admin-force overrides are available on Agency for one-off re-runs.

What if my Search Console connection breaks?

AI Ranking surfaces a clear humanized error like "Your Google Search Console connection has expired. Reconnect to continue." Your run state is preserved, no data is lost, and re-connecting through Settings restores access. Brands without GSC connected can still use the rest of the platform — AI Ranking is the only feature that requires it.

Find out what AI says about your real queries.

Start with a free Scorecard to see where you stand across four AI models. Then connect Search Console to run AI Ranking on your real buyer queries.

Check your visibility, free →
60 seconds · no credit card · no signup for first result