Two informational signals in the Mega Analyzer look at how a site handles two specific situations. A site that serves two distinct audiences. A site that ships AI products without saying how the AI is overseen. Both are optional, both are worth the engineering time when they apply.
Dual-persona navigation
If your site serves two audiences who want different things, you've got a decision to make. Option one: a single nav that tries to serve both and ends up serving neither. Option two: an explicit split at the top of the page, usually a "For X / For Y" toggle or a split-persona homepage.
Examples where option two measurably outperforms option one:
- Marketplaces. Buyers and sellers want different things. Etsy, Airbnb, and Zillow all run dual-persona navs on their homepages. The split is not cosmetic. Buyers want "browse products," sellers want "start selling" and "pricing." One nav can't hit both.
- Two-sided SaaS. A tool used by both developers and designers, or by both cities and builders, or by both investors and founders. Each side has different onboarding needs, different pricing relevance, different case studies.
- Professional services with two client types. A law firm that does both personal injury and M&A should not have a single "Services" dropdown that mixes them.
When it doesn't help:
- Single-audience sites trying to fake a second persona. "For Individuals / For Teams" on a product that has no team features is noise.
- Content sites. A blog for one audience doesn't need a nav split.
- Small service businesses with one customer type.
The audit flags it as informational because the question is "do you actually have two audiences." If you do, ship the split. If you don't, the extra nav is friction.
Interaction tip when you do ship it: persist the choice in localStorage so a returning visitor lands on the side they chose last time. Otherwise you're asking them to re-identify on every visit, which is annoying.
Human-in-the-loop and AI oversight messaging
This one is specifically for companies shipping AI products. Your enterprise buyers have a procurement checklist that includes "how is the AI supervised?" If your page doesn't answer that question, you've failed the checklist before a sales call ever happens.
What enterprise buyers want to see, in roughly this order:
- Explicit "human in the loop" or "human review" language where it's true. Spell out which outputs are reviewed by humans and which are generated directly.
- A description of how model outputs are audited. "Every output is reviewed by a licensed X before delivery" is concrete. "AI-augmented expertise" is marketing slop.
- Data-handling policy. What training data the model saw, whether customer data is used for training, how long data is retained. SOC 2 Type II or ISO 27001 certification if you've got it.
- An opt-out. If a customer wants to turn off the AI features, they should be able to, and your page should say so.
The opposite of this is the "we're AI-native" page that makes vague claims about accuracy without explaining how accuracy is measured. Enterprise buyers read that as a yellow flag. Regulators read it as a red flag. Neither is going to buy from you.
Two real examples of the pattern done well:
- A legal-tech product that says "Every document generated by our AI is reviewed by a licensed attorney in your jurisdiction before delivery" earns a lot of trust in the first read.
- A medical-imaging AI that lists its FDA clearance number, its radiologist-reviewer pool, and its false-positive rate up front earns enterprise sales in months instead of quarters.
Vague claims earn nothing.
Why the audit marks these informational
The dual-persona check only matters if you actually have two audiences. The human-in-the-loop check only matters if you actually ship AI. Neither is a universal fail. Both are strong recommendations in the cases where they apply, because getting them right removes a bigger friction than most sites realize they have.
The $97 Launch has a walkthrough of choosing between a single-audience nav and a dual-persona split, with before-and-after metrics from two sites that switched and one that tried to switch and rolled back.
Related reading
- E-E-A-T Analyzer, where both detections live
- Trust Signal Surface Audit, structural trust layer
- Claims and Attribution, the specific-claims side of AI-product trust
Fact-check notes and sources
- US FDA AI/ML-based Software as a Medical Device (SaMD) Action Plan, for the medical-imaging example.
- State bar association rules on unauthorized practice of law, for the legal-tech example.
- Baymard Institute research on persona-based navigation, for the dual-persona conversion findings.
This post is informational, not legal or regulatory advice. Specific examples are illustrative. Verify any compliance claims with your own counsel or regulatory body.
