Should I upload brand PDFs to an AI agent or let it research my brand automatically?

Upload your brand docs if you have them — it's faster, more accurate, and gives the invideo agent exactly the visual identity, tone, and guardrails you want enforced. Only let it research autonomously when you don't have those docs, or to supplement category/competitor context. The strongest setup is a hybrid: your owned guidelines first, agent research as a fill-in.

Upload, when you have brand guidelines ready. If you're the brand (or running ads for one), drop your own files into the invideo agent's context tab — visual guidelines, product catalogue, tone of voice, website, and a short treatment note covering camera language, lighting, composition, and a "standing don'ts" list (e.g. no plastic-looking fabric, no generic AI faces). One documented production seeded the agent with exactly three PDFs — Visual Guidelines, Product Catalogue, Treatment Note — and reused them across every ad in the project. Setup took 15–20 minutes once, then every subsequent ad inherited that context with zero re-prompting. As Hridaye, invideo's creative director, puts it: "If you're a brand that is running this for yourself, you don't have to ask the agent to research you. You can just upload your own brand guidelines deck with things like your visual identity, your tone, the kind of films you make, and so on."

Let it research, when you don't have docs — or to fill category context. The invideo agent can autonomously pull a brand's visual language, product details, and market positioning from the web and populate its own memory. In one production it analyzed Grüns end-to-end before scripting — surfacing the forest-green + yellow palette, the gummy-bear mascot, and the brand's identity cues without any uploaded deck. Use this when guidelines don't exist yet, when you're scaling ads for a client who hasn't sent assets, or to supplement your owned docs with competitor/category intelligence the agent can find but you haven't documented.

The hybrid is what most serious productions actually run. Upload everything you own — that locks the non-negotiables (logo, palette, tone, don'ts) to source-of-truth files. Then ask the invideo agent to research gaps: existing UGC ad formats on social, competitor pacing, market-specific cultural cues. One workflow explicitly tells the agent to study a brand's existing creator content before writing a single line of script — that's autonomous research stacked on top of uploaded brand context, not instead of it. The same logic applies if your guidelines are thin: jam with the invideo agent to generate the brief, lookbook, and treatment note from a conversation, then lock those as the persistent context.

Sharper inputs = sharper outputs — this is the real lever. A vague brief produces a vague agent; a tight brief makes the agent proactive and specific on every next step. Whatever route you take, the goal is the same: get structured, machine-readable context into the context tab once, so it persists across every image, video, and ad in that project. Static PDFs alone aren't magic — the agent needs the content of those docs absorbed into project memory, which is exactly what the context tab does. After that, you're not re-prompting brand rules per shot; you're directing creative.

Decision rule. Own the brand or have guidelines? Upload them — it's the fastest path to consistency and the only way to enforce your specific don'ts. Don't have guidelines, or want category/competitor context the agent finds online? Use autonomous research. For most real productions, do both: upload what you own, let the invideo agent research what you don't.

Watch some of these to see what works for you:

If you're a brand that is running this for yourself, you don't have to ask the agent to research you. You can just upload your own brand guidelines deck with things like your visual identity, your tone, the kind of films you make, and so on.

— Hridaye, invideo's creative director