Addability vs Findability: How AI Engines Just Outsourced Trust to Users

For twenty-five years, Google decided which sources you saw. In the past ten days, three AI engines started letting users decide instead.

On May 27, 2026, Google rolled Preferred Sources into AI Overviews and AI Mode globally — 345,000 sites are already user-selected, and selected sources get 2x click-through. On June 3, Perplexity launched its Personal Computer agent on Windows with hardened source-preference and domain-filter controls. On June 5-6, Microsoft shipped the Bing AI Search Choice extension to the Chrome Web Store, letting users pick which AI engine answers their queries at all.

Three engines. One direction. The trust signal AI engines once derived from their own ranking algorithms is being handed back to the user.

That changes the brand-side question. For two years, the GEO question has been "can engines find you?" Starting now, the harder question is "would a real human pick you?" And almost no measurement framework — including parts of Cited's own — captures this yet.

This piece is about that gap. What changed, what to call the new metric, what the May 2026 Cited Index data already shows about which Indian brands pass the bar, and what to do about it in the next week.

What Actually Happened in the Last Ten Days

Three engines, three different moves, same direction. Worth getting the specifics straight, because most coverage flattened them into "Google updated AI Search."

Google — Preferred Sources expansion into AI Overviews and AI Mode. Announced May 27, fully rolled out before the June 3 Search Console update that we covered last week. Users tap a "Preferred" badge in any AI Overview or AI Mode response to add that source to their personal trust list. Once added, future AI responses surface that source preferentially. 345,000 sites added globally by users themselves. Google's published data: 2x click-through for Preferred sources versus non-preferred citations in the same answer.

Perplexity — Focus mode + domain preferences hardened. The June 3 Personal Computer Windows launch came with Focus mode controls that let users limit retrieval to specific source types (web, academic, social, financial) and Search API-level domain filters that operate as user-set allowlists and denylists. The product narrative — "the answer engine you control" — is now sitting on top of an explicit source-preference layer.

Microsoft — Bing AI Search Choice extension (June 5-6). Jordi Ribas announced and shipped a Chrome extension that lets users choose which AI engine answers their queries from inside their default search experience. The framing is openly multi-engine: Microsoft is betting users want a choice of AI engine, not Google making the choice for them.

The pattern is what matters. For 25 years, source authority was an opaque algorithmic decision. Now three engines are putting a UI in front of it and asking the user to decide.

The New Question: Findability vs Addability

Every GEO tool on the market — Profound, Peec, Trakkr, Searchable, Cited's own GEO Score and Cited Index — measures findability. The audit questions are familiar:

• Can the engine crawl your site?
• Is your content extractable?
• Are you cited when a user prompts in your category?
• What's your share of voice against named competitors?

These are engine-side questions. They assume the engine's algorithm picks which sources to surface, and the brand's job is to be ready when the algorithm comes looking.

User-controlled source preferences change the question. They introduce addability: would a real human know your brand well enough to actively add it to their AI engine's settings panel?

Addability is human-side. It depends on:

• Brand recall. Will a user remember to type your name when prompted "add a site you trust"?
• Category association. When a user thinks "best resource on [your topic]," does your name surface?
• Cross-context consistency. Does the user encounter your brand frequently enough across email, social, Reddit, podcasts, conversation, and search to keep it top-of-mind?

These are the questions a CMO answered in 2005, but largely stopped answering after performance marketing took over budget allocation in the 2010s. They're back — wearing different clothes.

A brand can have a perfect GEO Score, get cited by AI engines today, and still fail addability. The engine's grounding pipeline can change tomorrow. The user-added trust signal is more durable than a citation that depends on whatever live web retrieval happened in the past three minutes.

What the Cited Index Already Reveals About Addability

We don't yet measure addability directly. But we measure a strong proxy: cross-engine consistency.

If a brand shows up as a top-3 mention across multiple AI engines that use different grounding pipelines, different training data, different citation behaviors — that brand has demonstrated enough independent signal strength that engines find them by multiple paths. It's the closest behavioural read on whether users would also find them.

Here's what the May 2026 Cited Index shows for Indian brands across ChatGPT, Gemini, Claude, and Perplexity. Top-3 brands per engine, per category, mapped to cross-engine presence:

The brands that pass cross-engine consistency aren't necessarily the brands with the best individual engine scores. Keka HR scores a perfect 100 on Claude — but its Perplexity score is 15. That's a textbook platform bias signal. It means Keka has built one strong source moat (likely Claude's training corpus + a specific grounding pattern) and depends on that single source.

Carlton, Wati, and Zimyo each have ~80% of their visibility riding on a single engine. If that engine retunes its retrieval — and they retune constantly — those brands disappear. They're invisible to addability already, because users encountering them on only one engine don't form the brand recall that would lead to a Preferred Sources entry.

Zoho CRM is the inverse. It appears as a top-3 brand on all four engines we track. Its visibility doesn't collapse if any single engine changes its stack. That cross-engine durability is functionally identical to the recall a user would need to add Zoho's domain to their preferences. In a world where addability matters, Zoho's position compounds. In the same world, single-engine concentration brands are structurally exposed.

How the 3-Layer Stack Updates

The 3-Layer AI Visibility Stack is Cited's mental model for AI visibility: Discoverability (can AI find you?), Citability (does AI use you?), Authority (does AI position you well?).

Findability lives in Layer 1 and Layer 2. Addability lives — newly — in Layer 3, but with a sharper edge than Authority alone.

Authority used to mean: third-party coverage that makes AI position you well when AI mentions you. Addability extends Authority to: third-party coverage and brand recall that makes a user position you well before AI is even involved.

That distinction matters for where the work happens.

• Layer 1 fixes (schema, crawl access, llms.txt) are days to weeks of engineering work.
• Layer 2 fixes (content structure, definitive statements, FAQ schema) are weeks to months of editorial work.
• Layer 3 addability work is six months to two years of brand-marketing work: email list activation, Reddit and Quora presence, Wikipedia, podcast appearances, conference talks, third-party review density, founder-led content that earns mentions.

Most Indian D2C brands have been told for two years that AI visibility is a Layer 1 + 2 problem. It isn't anymore. The brands that survive the next twelve months of engine-grounding changes are the brands that already passed the Layer 3 bar — and there are far fewer of them than the GEO industry has been pretending.

What the Cited 8 Doesn't Measure (Yet)

The Cited 8 — the eight metrics that power the Cited Index — captures cross-engine performance through Share of Voice (Metric 2) and AI Visibility Score (Metric 3). Both are strong findability metrics. Neither is a pure addability read.

A cleaner addability proxy would be a Cross-Engine Consistency Rate: the ratio between a brand's best-engine score and its worst-engine score. A brand at 100/15 (Keka HR's Claude/Perplexity gap) scores 0.15 on consistency. A brand at 95/85 scores ~0.89. The latter is the brand a user is more likely to add to Preferred Sources — because the user has encountered them everywhere.

This is data Cited already collects. We're working on whether it becomes a published metric in the next Cited Index iteration. Until it does, the closest stand-in is reading the platform bias map and ranking brands by how many engines they appear in as top-3, which is what the table above shows.

The shorter version: most GEO tools are still measuring 2024's question. The question for the next year is who's addable, not just who's findable.

What to Do This Week

Five moves for Indian brands, ordered by impact-per-hour:

1. Audit your cross-engine consistency. Pull your own visibility scores across ChatGPT, Gemini, Claude, and Perplexity. Compute best-engine ÷ worst-engine for your top 10 category prompts. A score under 0.3 means you have a single-engine concentration risk. Plan accordingly.

2. Activate your email list for Google Preferred Sources. If you have 10,000+ engaged subscribers or a comparable Instagram or WhatsApp community, write one note this week. Walk them through adding your site to Preferred Sources in Search settings. Each one earns you a compounding 2x click-through advantage on every future AI Overview citation you appear in.

3. Strengthen Layer 3 authority signals. Identify the top three places your category's customers gather online — Reddit threads, Quora topics, category review sites, niche newsletters, YouTube channels. Plan a six-month presence campaign on each. This is the slow work but it's the work that compounds.

4. Run the brand-recall test on a clean device. Open ChatGPT, Perplexity, Claude, and Gemini on a browser logged out of your account. Type the most common category prompt your customers use. Note which brands the engines surface. If your brand only surfaces on one of the four, you have an addability gap, not a measurement gap.

5. Re-prioritize content investment toward third-party citation density, not own-site optimization. The broader industry data on citation distribution shows 95.7% of AI citations on category prompts come from third-party sources, not the brand's own site. Your own site is a small slice of the visibility equation. Spend accordingly.

Where the Industry Goes from Here

User-controlled source preferences will spread. ChatGPT and Claude don't have an equivalent yet, but neither do they have a reason to ignore one. The pattern across Google, Perplexity, and Microsoft is too consistent to be coincidence — it's the regulatory pressure on grounding transparency colliding with user fatigue around opaque algorithmic ranking.

In the next twelve months, expect:

• An "addability score" category to emerge in GEO platforms. The first tool to publish a reliable cross-engine consistency metric anchored to user-preference behavior will own this conversation.
• Brand recall metrics resurfacing in CMO dashboards alongside performance metrics. Brand-as-noun matters again — and the brands that compounded into recall over a decade will see structural lift in AI visibility without changing anything technical.
• Sharper bifurcation between brands that built durable third-party citation density and brands that depended on a single engine's grounding stack. The latter group will look fine on today's GEO scans and disappear on tomorrow's engine update.

Cited's positioning across this shift is simple: we already track all four major engines, all eight metrics, and three layers of the AI visibility stack. The work ahead is to expose cross-engine consistency and addability proxies as first-class numbers — which is the next iteration of the Cited Index.

If you want to see where your brand sits on cross-engine consistency today, the Cited Index shows the full 202-brand May 2026 dataset across ChatGPT, Gemini, Claude, and Perplexity. The gap between your best-engine and worst-engine score is the addability read — and it's the number that's going to matter most over the next twelve months.