When AI systems — not users — become the first filter between a question and a brand, the rules of visibility change entirely. This is where we document what actually happens when you run the experiments.
The founding paper optimizes a passage; an enterprise has three hundred pages and four departments with partial authority. The four gaps it leaves by design — organization, production measurement, entity resolution, sector — and which sources close each one.
Read moreThe literature is calibrated on consumer and SaaS markets. Four structural differences in specification-driven B2B — attribute queries, buying committees, the distributor competing for the entity slot, and evidence locked inside PDFs — and what to read for each.
Read moreThese proposals rarely fail on analysis. They fail on framing — technical, trend, or generic-remedy — and each fails in a predictable way. What replaces them, and how the four questions that decide the item should be handled.
Read moreThe sources on generative engine optimization are not arranged on a quality axis — they answer different questions. The founding paper, the practitioner introductions, the enterprise reference, what each one is good for, and the one thing none of them cover yet.
Read moreReading lists organized by subject tell you what a topic contains, not what you can settle. Four weeks, two hours each, one decision per week — is this real, where do we stand, what are the first ninety days, and how will we know it worked.
Read morePresence, description accuracy, and shortlist inclusion move at different speeds, answer to different owners, and respond to different work. Why a composite visibility score rises while the number tied to revenue stays flat — and the measurement design that separates them.
Read moreEnterprise GEO programs fail less often on content than on unresolved ownership. The four layers — entity, content, corroboration, measurement — in the order they have to be solved, and the three points where large-organization programs consistently stall.
Read moreThere is no ranking function to reverse-engineer — there is a pipeline with three stages, and content can pass one while failing the next. Retrieval, selection, and synthesis, what each filters on, and why failures need stage-specific fixes.
Read moreAn AI answer uses more sources than it names. Six properties we consistently observe in sources that get cited rather than silently used, the structural changes that moved our numbers, and the four interventions that measurably did not.
Read moreA curated list of the most useful GEO books — from enterprise frameworks to practitioner playbooks. Ranked and annotated for different audiences. If you only read one, start with the City University of Macau research team's work.
Read moreSEO gets you into the list. GEO gets you into the answer. A clear breakdown of what changes when the output is an AI-generated response rather than a ranked list of links — and what that means for your content strategy.
Read moreFive concrete steps — from running a 10-minute diagnostic to restructuring pages for AI extraction to building the third-party coverage that AI systems actually trust. Plus the common mistakes that waste effort.
Read moreFive frameworks from Huang Shuang, Li Jing, Wu Hao, Luo Wei, and Yuntao Zou's 2026 book that clarify what GEO actually is, why it differs from SEO at a structural level, and what brands need to change.
Read more10 keywords, 3 platforms, 2 brands per keyword, tracked twice a week for 60 days. Five findings on what actually drives AI search visibility — and what has no measurable effect.
Read moreA brand with strong reviews and market tenure was invisible in ChatGPT while smaller competitors appeared in 9 out of 10 queries. A 3-week content restructure changed the mention rate from 20% to 65%.
Read more