How It Works

TopSlot is the AI Search Optimization Platform. Here is how we measure, fix, and optimize your AI visibility across ChatGPT, Claude, Gemini, and Perplexity. Unlike traditional SEO tools that track Google rankings, TopSlot specifically tests how AI models respond to real buyer-intent queries — either generated from your brand profile (Scorecard, Search Tracker) or pulled from your live Search Console data (AI Ranking). The result is either your AI Visibility Scorecard (a single composite score from 0 to 100) or a per-query citation map across four AI models. A 5-layer brand-safety classifier guarantees no query containing your brand, a competitor, or your domain ever reaches a model.

Auto-Detect Brand Information

Enter your brand name or website URL, and TopSlot automatically gathers key details about your business. We detect your industry, primary products or services, target audience, key competitors, and relevant keywords. This eliminates manual setup and ensures the audit is tailored to your brand from the start.

Generate Smart Prompts (or Pull Real GSC Queries via AI Ranking)

TopSlot ships two query-source paths. The default is Smart Prompts: based on detected brand information, the system generates a diverse set of natural-language prompts covering three core intent types:

• Buyer Discovery queries (“What is the best tool for X?”)
• Category queries (“Which services are available for Y?”)
• Decision Support queries (“How does Brand A compare to Brand B?”)

The alternative is AI Ranking: instead of generating prompts, the platform pulls every query that drove an organic impression to your site in the last 28 days from Google Search Console. Real buyer questions from real organic traffic, not synthetic guesses. Both paths feed the same downstream pipeline.

5-Layer Brand-Safety Classifier

Before any query reaches an AI model, it must pass a five-layer cascade. Layers run cheapest-first; the first rejection short-circuits and excludes the query. This is the architectural guarantee behind the “zero brand name in any query” methodology.

• Layer 1 — Word-boundary regex. Catches verbatim mentions of your brand, your competitors, or your domain stem.
• Layer 2 — Levenshtein ≤ 1. Catches typos and minor misspellings (e.g. “topslott”) on terms of length ≥ 4.
• Layer 3 — Batched GPT-4o classification. Sends batches of 25 queries to GPT-4o to catch semantic mentions the regex layers miss.
• Layer 4 — Pre-flight regex re-check. A final regex pass inside the model client immediately before the query is sent. Last line of defense.
• Layer 5 — Embedding cosine similarity ≥ 0.78. Compares the query embedding against a learned vector of your brand to catch descriptive references the prior layers can't see.

The classifier's steady-state expectation is zero Layer 4 emergency rejects. A daily Resend alert fires if any fire, signaling a coverage regression upstream.

Query AI Models

Each surviving query is sent to ChatGPT, Claude, Gemini, and Perplexity with web search enabled where the model supports it. The full response text is captured for every (query × model) leaf so it can be re-extracted if the scoring or labeling logic changes later.

Extract Mentions, Prominence, Citations, and Sentiment

For every AI-generated response, the extractor pulls four signal types:

• Mentions (M): whether your brand was named, anywhere in the response.
• Prominence (P): where your brand appeared. 100 points if first, 75 if top-3, 50 if top-5, 25 if mentioned at all, 0 if absent.
• Citations (C): whether your domain appeared as a link or URL in the response.
• Sentiment (S): the tone of the mention. Positive = 100, neutral = 50, negative = 0.
• Competitor presence: which tracked competitors appeared in the response and how they were positioned, recorded for every leaf so the ownership matrix can be rebuilt at any time.

Label Citation State (AI Ranking) or Calculate the Visibility Score

AI Ranking runs label every (query × model) leaf with one of four exclusive citation states:

• cited_with_url: your domain appears as a link or markdown URL in the response.
• name_only: your brand name appears verbatim, with no URL.
• descriptive_only: embedding cosine similarity ≥ 0.78 against your brand vector — semantically describing you without naming you.
• not_mentioned: none of the above.

Scorecard and Search Tracker runs compute the composite TopSlot Visibility Score and surface it as a category (Strong / Moderate / Emerging / Weak / Not visible) using a single locked formula: Score = M × 0.40 + P × 0.30 + C × 0.20 + S × 0.10. Mentions dominate because being named at all is the floor of visibility. Citations are the ceiling because they drive real referral traffic. Prominence governs whether you get clicked inside an AI summary. Sentiment is a tiebreaker, not a primary signal.

Generate Actionable Recommendations

Five prioritized recommendations are generated at the end of every paid run, each tagged to one of five categories:

• Schema: FAQPage, Organization, Article, Product schema gaps the pixel can deploy.
• Content: create / optimize / retire actions surfaced by the Content Engine's portfolio classifier.
• Citation: external sites that cite competitors but not you, and the press-release-shaped angle for each.
• Freshness: dateModified updates and content refreshes for pages AI models have started ignoring.
• Technical: robots.txt, llms.txt, and crawler-config fixes the pixel can ship without a developer.

Every recommendation is ranked by estimated impact on the next visibility delta and assigned to the system that ships the fix — Autopilot for technical/schema/freshness, the Content Engine for create/optimize/retire.

Execute Without a Developer

This is where most AI visibility tools stop. TopSlot does not. Install the Autopilot pixel on your site in 2 minutes (plus an optional NPM server adapter for Next.js, Express, Remix, or Nuxt to catch server-rendered fixes). Every recommendation flows through a deploy-approval workflow with a 200-entry audit log; once approved, the fix goes live, the verifier checks it every 10 minutes against your rendered HTML, and auto-rolls back regressions. Schema markup, llms.txt, robots.txt, FAQ pages, comparison content — all live, all versioned, all trackable. No developer required.