What exactly is AiVIS Cite Ledger?

AiVIS Cite Ledger is an evidence-backed AI citation audit and remediation system. It measures how AI models, ChatGPT, Perplexity, Claude, and Google AI, interpret, trust, and cite your web content. The result is a 0–100 CITE LEDGER score across seven weighted dimensions, with BRAG evidence-linked findings and prioritized fixes.

How is this different from Semrush, Ahrefs, or Clearscope?

Traditional SEO tools optimize for keyword ranking in traditional search results. AiVIS Cite Ledger audits citation readiness, the specific signals that determine whether AI answer engines retrieve and cite your content. A page can rank #1 on Google and be completely invisible to AI answer synthesis.

What is BRAG and how does it make findings trustworthy?

BRAG stands for Based-Retrieval-Auditable-Grading. It ties every audit finding to a real, traceable element on your page, a specific heading, schema block, meta tag, or content section, each carrying a unique evidence ID. You can open the element, verify the finding, and recheck it after a fix. No black boxes.

Is the free tier genuinely useful, or just a teaser?

The Observer (free) plan gives you a full composite CITE LEDGER score and all seven dimension letter grades, so you can see exactly where you stand and which categories are hurting you. BRAG evidence IDs and prioritized fix packs require a paid plan. Three audits per month, no credit card.

Your Brand Mention Count Is Lying to You About AI Visibility | AiVIS Cite Ledger Blogs

By R. Mason · 2026-04-15 · 9 min read · AEO

Mention volume is a vanity metric. AI models weight sources by credibility, one Wikipedia reference outweighs hundreds of directory listings. See how Mention Juice Score measures what actually drives AI citation strength.

Key Takeaways

Why mention volume is a misleading metric for AI visibility
How source credibility weights are assigned across 19 platforms
The spam and duplicate filtering logic that removes false signal
How sentiment and recency multipliers affect your final score
Practical Reddit strategy for generating high-credibility mentions

Article

Why mention counting gets AI visibility wrong

Brand mention tracking tools report volume. They count how many times your brand name appeared online this week. That number is widely reported, tracked in dashboards, and treated as a proxy for brand presence.

It is the wrong metric for AI visibility.

AI models do not count mentions. They weight them. A single thread on Hacker News where your product is discussed in technical depth carries more weight in model training data than 500 low-effort product listings on directories that were created in bulk. A Wikipedia article that references your brand carries more weight than 200 tweets. Three thoughtful Stack Overflow answers that mention your tool carry more weight than a page-three Google News story.

The difference matters because AI models are trained and tuned on corpora where source quality is implicitly encoded through signals like PageRank, domain authority, editorial standards, and community engagement. Sources that consistently produce high-quality signal accumulate influence in model training, and that influence transfers to citation decisions at inference time.

How the Mention Juice Score works

The Mention Juice Score assigns a credibility weight to each source where a brand mention can occur, then applies a recency multiplier and a sentiment multiplier to compute a weighted score. The formula is:

```

mention_juice = Σ(weight × sentiment_mult × recency_mult) / max_possible × 100

```

A volume bonus (log₂ scale, max +15 points) rewards brands with broad legitimate coverage without allowing volume to dominate quality.

Source credibility tiers

The 19 sources tracked are grouped into credibility tiers based on their historical influence as training data signal sources:

**Tier 1, Maximum credibility (weight 0.80–1.0)**

Wikipedia (1.0), the highest-weight source. Wikipedia articles directly influence knowledge graph entries and are heavily represented in model training corpora.
Reddit (0.85), co

Enable JavaScript for the full interactive reading experience with related articles and discussion.