Agent successfully gathered the core information (what, who, pricing, differentiation) from three direct fetches of the site. However, pricing and competitive positioning remain opaque: the site does not name competitors, omits annual pricing, and obscures per-seat usage allowances, requiring the user to contact sales or infer from sparse model rate disclosures.

• ›Pricing page is detailed but incomplete: usage is metered per model at unpublished rates and pooled by team, making cost prediction difficult. The site withholds per-seat dollar amounts, forcing users to contact sales or estimate in-app.
• ›No competitive comparison exists on the site — the agent inferred differentiation from implicit claims (model recency, single canvas, no tool-switching) rather than head-to-head comparison, so positioning vs. Adobe Firefly, Figma, or ComfyUI is a gap.
• ›Certification claims conflict between pages: homepage states 'SOC 2 Type 1 & 2' but the enterprise FAQ clarifies Type 1 only, with Type 2 and IP indemnification pending Q3 2026. Trust documentation was not fetched.
• ›Site navigation surfaces pricing and enterprise pages directly from the homepage, making those resources easily discoverable; however, feature documentation (Techniques, Studios, Actions) lacks a single consolidated explainer — naming overlaps and terminology requires inference.
• ›Free tier quota is model-specific ('17 generations' at $0.151 per Nano Banana 2 image), not a general allowance, making it confusing as a marketing statement and non-portable across model choices.