Andi Review

No pricing page, no API docs, no changelog, no known funding. That's four red flags before I even look at the product. Perplexity AI raised $165M and is shipping weekly. That's who you compare Andi against, and that comparison doesn't go well on viability.

The product itself is solid for what it is. Zero tracking, no cookies, ad-free by architecture, and per-result summarization without leaving the interface — those aren't marketing claims, they're structural. The Chrome extension and PWA suggest real shipping cadence. But it's a free consumer tool with no stated paid tier, which means no revenue model I can see.

I won't take this to the board as a vendor relationship. The tradeoff is real: strong product execution, completely unclear business sustainability. Useful for a privacy-conscious employee who wants a Google alternative. Not a procurement conversation.

The Trantora index scoring 450+ data points per page shows real investment in content quality signals — that's meaningful. Per-result summarization and the Explain feature address real friction in research workflows. For an individual knowledge worker doing open-web research, this is a cleaner experience than Google's ad-cluttered results pages.

The gap is structural. No API access for internal knowledge bases, no user-level search history, no glossary or taxonomy controls, no connector layer to enterprise repositories. Perplexity AI ships team plans with shared spaces and source pinning — Andi has none of that. The zero-tracking architecture that's a privacy strength actively prevents the query-pattern analysis that makes enterprise knowledge systems improve over time.

If we adopt this for an organizational knowledge program, in 3 years we have a tool that's still optimized for anonymous consumer queries with no institutional memory built on top. Free pricing with no paid tier means no roadmap leverage and unclear sustainability.

$0/seat × 50 users × 36 months = $0 TCO. Hard to beat on paper. No SSO tax, no overage rate, no per-seat negotiation. Andi's pricing page doesn't exist because there's nothing to price. That's either a gift or a warning.

The 15-billion-page proprietary index with 450+ credibility signals per page is a real architectural differentiator versus Perplexity AI, which leans on third-party APIs. Zero-cookie, no-IP-storage privacy architecture adds procurement value for privacy-sensitive orgs. Chrome extension and PWA cover the deployment surface adequately.

The tradeoff: no paid tier means no SLA, no support contract, no vendor accountability. Year 3 TCO is $0 — until it isn't. Free products pivot. Procurement can't negotiate what has no price. For teams needing contractual guarantees, Perplexity's $20/seat business tier buys you an actual agreement. Andi buys you nothing to sign, which cuts both ways.

The zero-cookie, no-IP-logging architecture is real — not a privacy policy promise but a structural design choice. For researchers who worry about institutional queries leaking through Google's ad targeting, that matters. The per-result Summarize button generating ~200-word digests is actually useful for triage: scan ten sources without ten tabs. The Explain feature on dense results is a quiet win for interdisciplinary reading.

Day three is where the gaps show. No changelog, no docs, no API visibility for public users — the capability flags all read N. Perplexity gives you follow-up threading, citation confidence signals, and a focused Academic mode. Andi doesn't expose query refinement depth or session continuity. Researching across a week of related queries means rebuilding context every session.

The proprietary Trantora index scoring 450+ credibility signals per page is the most interesting architectural claim here. But without a public blog or transparency report, there's no way to audit index quality. For exploratory research where source credibility is the whole game, that's a real gap.

The pitch is dead simple. Type a question in plain language, get an answer with sources on the right so you can verify. No blue-link wall, no sponsored garbage, no cookie banner. The per-result Summarize button and the Explain feature are the kind of small things that make you wonder why Google never bothered. The proprietary Trantora index at 15 billion pages means they're not just wrapping someone else's API and calling it a product.

That said, no changelog, no blog, no pricing page — the evidence trail is thin. Compared to Perplexity, which has a clear Pro tier and active public development, Andi feels quieter. Harder to know where it's going. Free-forever is great until it isn't, and there's no public funding data to reassure you.

The PWA covers mobile, which is better than nothing, but category norm says a real native app matters. Day three you'll have a strong opinion about whether the card layout actually helps you think faster. I suspect most people will.

Three flags before I go deeper. One: no company name visible, no funding data, no changelog. Two: the capabilities page shows API=N, blog=N, pricing-page=N — that's a thin public footprint for a product competing against Microsoft Copilot. Three: 'AI search for the next generation' is exactly the kind of tagline that ages poorly.

What's actually credible: the Trantora proprietary index — 15 billion pages, 450+ scored data points per page — is a real differentiator if it holds up. Per-result summarization, the ad blocker, zero-cookie architecture, and a Chrome extension are all legitimate features. The privacy story is coherent, not bolted-on marketing.

The exit story is fine. It's a web app, no account required, no data stored. You walk away clean. But long-term viability worries me. Entirely free, no paid tier, no public investors — someone's paying for those LLM calls. Perplexity charges $20/month and still burns cash. This math doesn't close without a funding story I can't find.