AdForge

AI Output Disclaimer

Last updated: May 17, 2026

AdForge Studios uses third-party artificial-intelligence models to generate text, images, intelligence analysis, and simulated audience feedback. This page explains what those outputs are, what they are not, and what you must do before relying on them. It is part of our Terms of Service.

1. AI outputs are probabilistic, not authoritative

The platform uses large language models (Anthropic Claude via OpenRouter) for text outputs and Google Gemini 3.1 Flash Image (Nano Banana 2) for image outputs. These are statistical models. They produce outputs that look authoritative but can contain:

  • Factual errors and fabricated statistics or quotes
  • Outdated information that no longer reflects the world
  • Biased framing or omissions
  • Unintentional resemblance to copyrighted or trademarked third-party works
  • Confidently-stated claims that are simply wrong

AdForge does not warrant the accuracy, completeness, or fitness-for-purpose of any AI output.

2. You must review outputs before using them

You are solely responsible for reviewing, fact-checking, and approving every AI-generated output before:

  • Publishing it to any ad platform (Meta, Google, LinkedIn, TikTok, etc.)
  • Sharing it with a client or stakeholder as research or a deliverable
  • Using it in any public-facing channel
  • Making a business decision based on it

3. Intelligence reports — what they are

  • Intelligence reports are AI summaries of publicly-available data scraped from third-party websites (TikTok, Instagram, YouTube, Reddit, Amazon, Walmart, Target, Trustpilot, Yelp, Google Maps, Google Shopping, Pinterest, LinkedIn, X / Twitter, Meta Ad Library, TikTok Shop, Facebook).
  • Recency, completeness, and accuracy of source data are best-effort, not guaranteed. Scrapers can fail, rate-limit, or return partial data. The model can misinterpret what it sees.
  • Reports are produced under server-side spend caps. If a report hits the cap or a scraper fails the report is marked partial and the missing-data state is reflected in the analysis. Partial reports are still billed.
  • Reports are a starting point for human research and creative direction, not a replacement for primary consumer research, market sizing studies, or legal due diligence.

4. Generated personas — what they are

Personas the platform generates are statistical composites built from public data and AI synthesis. They are not real people, not survey respondents, and not a substitute for researching your actual customer base. Names, photos, quotes, and biographical details in a persona are generated illustrations, not facts about any specific individual.

5. Design Review feedback — what it is

Design Reviews simulate persona-style critique of a creative. They are not human focus groups and should not be cited as such. The responses come from an LLM playing a role; the LLM does not have access to a population of real consumers. Use Design Review for first-draft direction-checking and to accelerate internal review — not as primary consumer research.

6. Image creatives — IP & brand safety

  • AI-generated images may inadvertently resemble copyrighted works, trademarks, or real people. You are responsible for IP clearance before commercial use. AdForge does not provide an IP indemnity for AI-generated content.
  • The platform includes brand-safety aids (for example the unusable_for_creative flag on certain intelligence-report pain points) to reduce the chance of generating obviously off-brand or off-topic creative. These aids are quality helpers, not guarantees.
  • Generation works by prompting the model with your brand context. The model decides exactly what the image looks like. Brand color enforcement, logo placement, and on-image text rendering are best-effort — treat the image as a draft that requires review.

7. AI outputs are not professional advice

Do not use AI outputs as the basis for, or a substitute for, professional advice in:

  • Medicine, mental health, nutrition, fitness, or any health-related decision
  • Law, regulatory compliance, tax, accounting
  • Finance, investments, insurance, lending
  • Hiring, eligibility, credit, housing, or other consequential decisions about individuals

8. Marketing language vs contract

Marketing language such as "agency-grade," "deep research," "essential," "research-backed," "in seconds," or comparable phrases describes a quality target and a feature tier. It is not a contractual guarantee of accuracy, performance, or business outcome. Binding contractual terms live in the Terms of Service.

9. Changes

AI providers and behavior evolve quickly. We update this page when our pipeline changes materially.