For a human reader, quality means engaging, well-written content that is useful and accurate. For an AI inference engine evaluating citation candidates, quality means something more specific: factual accuracy, structural clarity, extractability, and comprehensive coverage of the topic without gaps or contradictions.
Factual accuracy matters because inference engines apply consistency checks — claims that contradict established knowledge trigger skepticism signals. A page that accurately describes the cost, process, and timing of soft wash roof cleaning in Florida is more citable than a page that makes imprecise or inaccurate claims about the same topic.
A core quality signal that is specific to AI citation contexts is extractability — how cleanly the relevant information can be isolated from the page and presented as a citation. Declarative factual statements in direct sentence structure are highly extractable. The same information embedded in marketing copy with hedging qualifications is not.
"Soft wash roof cleaning costs between $200 and $600 for a standard Sarasota home depending on roof size and contamination level" is extractable. "Our competitively priced soft wash services deliver exceptional value across a wide range of roof cleaning situations" is not extractable — it contains no citable fact.
Volume stops helping — and starts hurting — when additional pages dilute the quality standard. A domain where 400 of 450 pages are high-quality and 50 are thin or repetitive has a lower average quality signal than a domain where all 400 pages meet the quality standard.
Inference engines evaluate domains holistically. A domain where some pages produce skepticism signals can see those signals affect citation probability for otherwise-high-quality pages on the same domain. Maintaining quality standards across the full page inventory is the ongoing discipline that determines whether volume produces compounding authority or compounding dilution.