The Pondral journal.

Every scored engine call carried a finance-analyst persona left over from an earlier product. A 324-call paired test measured the cost: +1.04 points overall, +14.31 on open-ended advisory questions across 9 paired comparisons, -1.44 on the finance controls. PROXY throughout, measured under methodology v2.0.3.

An agency that sells Answer Engine Optimization scores 95.3 when asked about by name and 15.6 on the discovery questions its buyers actually type. Five engines. Correction, August 2026: the run counts originally published here are unverified. We do not name teardown subjects.

Two formulas existed. Three production bugs surfaced. Our generic visibility was zero. The full audit trail of our methodology reconciliation.

Great branding and SEO get you considered by AI assistants, not necessarily named. What the AI Visibility score measures, and where a strong brand still falls short.

Retrieval vs generation, why citations matter, and Competitive Presence. The mechanics behind AI-assistant recommendations.

Five published signals. How Pondral scores presence, prominence, context, citation links, and competitive presence, and why the weights are public.

A top-ranked national spa franchise scores 38.7/100: strong on branded questions, weak on the discovery questions new customers ask.

B2B buyers research vendors with ChatGPT and Perplexity before they call sales. Five things to work on if your brand isn't named in those answers.

Training data vs. live search, entity signals, and authority hierarchy: how ChatGPT, Claude, Gemini, Perplexity, and Grok decide which brands appear.

Five factors, each weighted, each public. Why they're weighted the way they are, and what edge cases we've learned to handle.

Your About page is where AI engines learn who you are and what category you belong in. Here's what to put on it.

Structured data is how you describe yourself to AI before it gets a question about you. Here's the schema that matters.

Correction: the ninety-day study described in this post was never run. Our first measurement data of any kind dates from 2 May 2026, the day after this was published.

People still look for brands. They just don't start with Google. Here's what that shift means for your visibility strategy.

Why single-run scores have limits, what inter-rater reliability tools like Cohen's kappa offer, and what Pondral measures today.

How AI engines decide which brands to cite, how to measure your presence, and what to do about the gaps.

Ten well-known brands and the public patterns they share: content depth, third-party coverage, entity consistency. Illustrative examples, not a measured study.

Each engine has different citation habits. Here's what that means for your brand's visibility strategy.

The design decisions behind the five-factor rubric: what we built, what we changed, and what we got wrong.

Structured data lets AI engines parse your brand before they answer a question about it. The playbook, with examples.

How AI assistants pick products to recommend, and what an e-commerce brand can give them to quote.

Five changes you can make this week, ordered by effort.

AEO is the practice of optimizing for answers, not rankings. Here's the definition that actually holds up.

Content quality still matters. What carries over from SEO to AI answers, and what is new.

A working vocabulary: AI visibility, brand SOV, citation link, entity consistency, and 20-plus more terms defined.