Setting Up Foglift with an AI Assistant

1. Token Budget Math

Every AI Visibility scan costs tokens. The cost depends on which engines are enabled. Before creating prompts, calculate how many the customer's plan can sustain. Free accounts get weekly automated Google AI Overview scans, plus manual checks any time. Paid plans add ChatGPT, Perplexity, Gemini, Claude, and faster plan cadences: Launch daily, Growth twice daily, Enterprise hourly.

Per-Engine Token Costs

Each prompt is sent to every enabled engine on each scan. The total cost per prompt per scan is the sum of all enabled engine costs.

Monthly Token Limits by Plan

The Formula

Calculate the maximum sustainable prompts per brand before creating anything.

prompts_per_brand = monthly_tokens / (brand_count × scans_per_month × cost_per_prompt)

Where:
  monthly_tokens   = your plan's token limit (see table above)
  brand_count      = number of brands being tracked
  scans_per_month  = scan frequency: weekly ≈ 4.3, daily ≈ 30, monthly = 1
  cost_per_prompt  = sum of enabled engine costs (17 if all 5 enabled)

Worked Examples (All 5 Engines Enabled)

Find your plan row and scan frequency column. The number is the maximum prompts per brand.

Highlighted row: Launch daily, 1 brand, all 5 engines = 7 prompts per brand. Launch daily, 3 brands, all 5 engines = 2 prompts per brand, so an assistant should pause and ask whether to reduce engine coverage, track fewer brands, slow cadence with explicit confirmation, or upgrade. Values shown as 0 mean the budget cannot sustain even 1 prompt at that cadence.

2. What Foglift Tracks via Prompts

Foglift categorizes every prompt into one of four prompt categories, not AI Visibility Ladder tiers. The full AI Visibility Ladder has six tiers; these categories feed the prompt-driven visibility signals. Each category measures a different kind of search behavior. For meaningful scores, create at least 2 prompts per category per brand.

Minimum total: 8 prompts per brand (2 per category). Recommended: 12 to 20 per brand when the confirmed cadence and engine mix can sustain it.

Anti-Patterns to Avoid

These prompt styles look relevant but produce noisy or useless data.

• Too generic: "Tell me about software" will not reliably surface any specific brand.
• Too specific: "What is [BRAND]'s pricing for the Growth tier in Q2 2026?" tests LLM knowledge freshness, not brand visibility.
• Off-category: "Best pizza restaurants in Denver" for a SaaS company. Industry prompts should be within your actual market vertical.
• Duplicate intents: "What is [BRAND]?" and "Explain [BRAND]" test the same thing. Use the second slot for a different angle like "Who founded [BRAND]?"
• Competitor-only: "What is [COMPETITOR]?" does not tell you anything about your own visibility. Put competitor names in comparison prompts, where the head-to-head framing matters.

Example Prompt Mix (18 prompts)

For a brand called "Acme" in the project management category, competing with Asana and Monday.com:

Brand (5 prompts):
  "What is Acme?"
  "Tell me about Acme project management"
  "Who founded Acme?"
  "Is Acme a good project management tool?"
  "What does Acme do?"

Comparison (5 prompts):
  "Acme vs Asana"
  "Acme vs Monday.com"
  "Acme alternatives"
  "Asana vs Monday.com vs Acme"
  "Acme compared to Notion for project management"

Review (4 prompts):
  "Best project management tools 2026"
  "Acme reviews"
  "Top project management software for startups"
  "Best tools for managing remote teams"

Industry (4 prompts):
  "How to manage a software project"
  "Project management tools for small teams"
  "How to track tasks across remote teams"
  "Best way to organize a product roadmap"

3. Interactive Setup Walkthrough

If you are an AI assistant helping a user set up Foglift, follow this sequence. Ask each question, wait for the answer, and use the responses to compute the right configuration.

POST /api/v1/prompts
Content-Type: application/json
X-API-Key: sk_fog_...

{
  "prompt": "What is Acme?",
  "brand_id": "brand-uuid-here"
}

4. Installing the AI Traffic Tracker

The @foglift/tracker package detects two things: AI bots crawling your site (GPTBot, ClaudeBot, PerplexityBot) and humans clicking through from AI search results (ChatGPT, Claude, Perplexity, Gemini, Copilot). No cookies, no fingerprinting, no PII collected.

Next.js (Recommended)

npm install @foglift/tracker

import { trackAITraffic } from "@foglift/tracker/nextjs";

export const middleware = trackAITraffic({
  apiKey: process.env.FOGLIFT_API_KEY!,
});

export const config = {
  matcher: ["/((?!_next/static|_next/image|favicon.ico).*)"],
};

If the user already has a middleware file, compose with the existing middleware:

import { trackAITraffic } from "@foglift/tracker/nextjs";
import { existingMiddleware } from "./my-middleware";

export const middleware = trackAITraffic({
  apiKey: process.env.FOGLIFT_API_KEY!,
  next: existingMiddleware,
});

Multi-brand SDK installs on @foglift/tracker v1.3.0+ should pass brandId with the brand UUID from Dashboard → Developer → Your brands. The SDK sends it as brand_id; single-brand installs can omit it. The v1.3.0 npm publish is waiting on the 2FA-gated release step.

Express

import express from "express";
import { trackAITraffic } from "@foglift/tracker/express";

const app = express();
app.use(trackAITraffic({ apiKey: process.env.FOGLIFT_API_KEY! }));

Plain Node.js

import { createAITrafficTracker } from "@foglift/tracker/node";

const tracker = createAITrafficTracker({
  apiKey: process.env.FOGLIFT_API_KEY!,
});

// Call tracker.track(req) in your request handler
tracker.track(req);

Script Tag (WordPress, Webflow, Wix, Static Sites)

For platforms without server-side access, paste this in your site's <head> section. It detects AI engine referrals only (bot detection requires server-side access).

<script async
  src="https://foglift.io/pixel.js"
  data-foglift-key="sk_fog_..."
  data-foglift-brand="BRAND_UUID">
</script>

The pixel is under 3KB. It reads document.referrer on page load, checks if the visitor came from an AI engine, and fires a single POST. Use the brand UUID from Dashboard → Developer → Your brands for data-foglift-brand. No cookies, no fingerprinting.

Where to Find the API Key

Generate an API key at Dashboard → Developer. Keys start with sk_fog_ and should be stored as an environment variable, never committed to source control.

After installation, AI traffic data appears in your Crawler Analytics dashboard within seconds of the first matching visit. AI assistants can also call the MCP tool get_referrer_analytics with brand_id and days to summarize total visits, engine breakdown, top paths, and daily trend. For setup, call recommend_setup first, then apply_setup_recommendation with the target brand_id to create the categorized prompt set. For optimization loops, call get_recommendations to fetch diagnostics and a reasoning recipe, use your agent to reason over the fix, then call submit_recommendation so Foglift stores it as customer_agent. After apply_recommendation, re-run run_ai_visibility and get_ai_results to verify movement.

5. Connecting Google Search Console

Google Search Console integration adds traditional search performance data (clicks, impressions, CTR, average position) alongside your AI visibility metrics. It is optional but recommended for a complete picture.

What Foglift does with GSC data: Correlates traditional search queries with AI visibility data to identify content gaps. For example, if a query drives strong Google traffic but your brand is absent from ChatGPT answers for the same query, Foglift surfaces that as an optimization opportunity.

Privacy: GSC tokens are stored server-side, scoped to your account. Team members can view data for brand domains only. Foglift never writes to or modifies your Search Console settings.

Frequently Asked Questions