Skip to main content

​
Overview

The Analysis (Full Report) goes much deeper and takes up to 90 seconds. It consists of five steps executed sequentially.

​
Three Analysis Strands

The analysis is based on three parallel or sequential strands:

​
1. Technical Audit (Context/Clarity) – domains only

Goal: How well is your website readable and understandable for AI models?
  • Website crawl – Your website is crawled, important pages analyzed
  • Structured data – Schema.org/JSON-LD, meta tags, heading hierarchy, Open Graph
  • llms.txt – Is it present? Does it offer AI models a structured summary?
  • robots.txt – Which AI bots (GPTBot, ClaudeBot, PerplexityBot, etc.) are allowed to crawl?
  • Image alt texts, canonical, meta robots, HTTPS, response time
Result: The Clarity score (0–100) shows how well AI can capture and correctly assign your content. For person scans this strand is skipped.

​
2. Visibility (Unseeded)

Goal: Are you recommended in general queries too – without your brand being explicitly mentioned?
  • Search queries are generated based on industry and offering (navigational, commercial, comparative, optional custom)
  • Unseeded = The AI receives no hint about your brand. This tests whether you are organically recommended – as in a real customer query
  • Search models: Perplexity Sonar and GPT-4o-mini Search
  • Per intent it is checked: Is your brand mentioned? At which position?
  • Position scoring: Position 0 = highest score, Position 3+ = lower score
Result: The Presence score (0–100) shows how often and prominently you appear in AI search results. For person scans, recognition rate from the Recognition strand is additionally included (50/50 blend).

​
3. Recognition & Risk

Goal: Do models know you? Are there mix-ups or misinformation?
  • Knowledge mode: Several LLMs are directly asked – e.g. “Do you know [brand]?” – without web search
  • Search mode: Same models with web search to include current sources
  • Per model: Confidence (how certain?), external sources, ambiguities (mix-ups)
  • Aggregation: Recognition rate, average confidence, external authority
Result: Recognition score and Risk score – contradictions and uncertainties flow into the Risk score.

​
Scoring (Full Report)

After the three strands, scores are calculated:
DimensionSource
RecognitionRecognition model results, known_ratio, external_authority_score
PresenceUnseeded intent results; for persons: 50% Unseeded + 50% recognition rate
ClarityTechnical audit (domain only); person scans: 0
RiskRecognition aggregation (ambiguities, uncertainties)
Weighting for domains:
  • Recognition: 25%
  • Presence: 35%
  • Clarity: 30%
  • Risk: 10%
Weighting for persons:
  • Recognition: 60%
  • Presence: 15%
  • Clarity: 0% (not applicable)
  • Risk: 25%

​
Action Recommendations

Action recommendations are generated using Flagship models. They draw on the three analysis strands (technical audit, visibility, recognition) and apply best practices to produce prioritized measures:
  • Each recommendation refers to a finding source (technical audit, visibility analysis, recognition test, score engine)
  • E-E-A-T assignment: Experience, Expertise, Authoritativeness, Trustworthiness
  • Priority: High, Medium, Low
  • Effort: Quick Win, Medium, Strategic
  • Expected effect is described

​
Temporal Sequence

  1. Context-Clarity (technical audit) – for domains
  2. Unseeded (visibility) – can run parallel with Recognition
  3. Recognition/Risk
  4. Scoring – calculation of final scores
  5. Actions – generation of action recommendations
Timeout for the full analysis: approx. 95 seconds.

​
Entity Types

TypeDescriptionClarityNote
DomainWebsite (e.g. makemerank.ai)YesTechnical audit is performed
PersonName + topic (e.g. “John Smith, B2B SEO”)NoSpecify topic for more precise Presence scoring

​
LinkedIn Context

When you have a LinkedIn profile connected (or pass a LinkedIn URL), context from your profile can optionally be used to build more specific queries. Instead of generic formulations like “SEO consultant”, the Primary Category and Brand Claim are derived automatically from your actual profile. This makes the Intent and Recognition calls more precise – and yields more accurate results. MakeMeRank uses the official LinkedIn API with read-only access (no risk). Only publicly available data is used – e.g. headline, summary, and posts. Nothing private is accessed.

​
Result Variability

AI models respond non-deterministically. With repeated scans, scores can vary slightly. Broad trends stay stable; small deviations (e.g. ±5 points) are normal. For this reason, some calls are executed multiple times in sequence and the results are averaged. In addition, the API calls are configured to return as consistent and realistic results as possible (e.g. via temperature settings and prompt design).