Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.adloft.ai/llms.txt

Use this file to discover all available pages before exploring further.

When you generate, AdLoft runs a durable background workflow of seven specialized AI agents. They run in order, each building on the last, and each explains its reasoning. You can leave the page — generation continues in the background.

​
The seven agents

1

Page Analyzer

Reads your landing page and extracts the offer, target audience, value props, recommended angles, and CTAs.
2

Structure

Designs the campaign structure — the campaign name, ad set layout, and a hypothesis for each ad set.
3

Targeting

Builds the audience for each ad set — locations, age, gender, interests, behaviors, and placements — favoring your strongest historical audiences.
4

Creative

Selects the best creative assets per ad set from your existing library, favoring proven winners and landing-page fit.
5

Copy

Writes the ads — primary text, headlines, descriptions, and CTAs — matched to each ad set’s hypothesis.
6

Budget

Allocates your total budget across the ad sets based on your strategy and the signal it has.
7

Director

Makes the final campaign-level decisions — final name, bid strategy, and overall rationale.
After the agents finish, AdLoft assembles their outputs into one campaign plan and materializes it into an editable draft, validating that it’s launch-ready against Meta.

​
What the AI sees

The agents work from:
  • Your campaign inputs (objective, budget, identity, strategies).
  • The landing page’s title and text.
  • Up to two dozen creative assets from the brand’s library.
  • Your top historical creatives, audiences, and campaigns for the chosen lookback window.
The Creative agent only selects from creative assets that already exist in your brand. AI Launch assembles and recommends — it doesn’t generate brand-new images or video.

​
Reliability

  • Generation runs in the background and survives you leaving the page.
  • Each step is processed once and stored with its rationale.
  • If a step stalls for too long, the run is marked interrupted and can be resumed from the last completed step. See Runs & results.

​
Explainability

Every agent records its rationale and any exclusions. These power the live “thinking” line during generation and the AI strategy summary in the finished draft, so you always understand why the AI made each choice.