We Audited 50 D2C Brands for AI Visibility. Most Were Invisible

We ran an experiment we had been putting off. For fifty Indian direct-to-consumer brands, we sat down and asked the AI assistants the questions their buyers actually ask - not the brand names, but the category questions a shopper types before they have decided who to buy from. "Best clean-label protein powder in India." "Which sunscreen is good for oily skin under a helmet." "Affordable cold-pressed oils delivered in Mumbai." For each brand we assembled a battery of these questions, ran them across ChatGPT, Perplexity and Gemini, and recorded one thing: did the brand ever surface.

The result was worse than we expected, and we expected it to be bad. Thirty-eight of the fifty brands - seventy-six percent - never appeared once, in any answer, on any engine, for any of their own category questions. They were, in the only sense that matters to an AI-first buyer, invisible. Nine surfaced occasionally. Three showed up reliably. And the brands that were invisible were not obscure. Several of them rank on page one of Google for their head terms and spend real money on performance media every month. They have simply been optimising for a search surface that a growing share of their buyers no longer use first.

This piece is the full audit, written the way we ran it: the method, the raw pattern in the scores, the five reasons brands were invisible, what the three visible outliers did differently, and the scoring framework we built out of it - the one we now use to decide whether any brand is what we internally call AI-recommendable. If you run a D2C brand, this is a mirror. Most of what it reflects is fixable, and cheaper to fix than you think.

Why We Ran This Audit At All

We did not run this to produce a scary statistic. We ran it because we kept seeing the same disconnect in client conversations, and we wanted to measure it rather than assert it.

The disconnect is this. Brands know their Google rankings. They watch positions, impressions, clicks; they have dashboards for it. Almost none of them knew their AI visibility, because there is no default dashboard for it, and because for years there was no reason to care. But the buyer journey has quietly forked. A large and rising share of category research now begins inside an assistant or on an AI answer surface, a shift we track in detail in our AI Search Statistics 2026 reference. When a shopper asks ChatGPT "what's a good affordable D2C sunscreen for Indian skin" and gets three names back, being ranked fourth on Google for "best sunscreen india" does nothing. You were not in the room where the recommendation happened.

We had written about the shape of this problem before - the conceptual version in Ranking on Google but Missing on ChatGPT, and the category-specific version in What Indian D2C Brands Get Wrong About SEO in the AI Era. What we had never done was quantify it across a real sample. So we did. Fifty brands is not a census, and we are not claiming statistical authority over the whole D2C market. But fifty brands, audited with a consistent method, is enough to see the pattern clearly - and the pattern was unambiguous.

The Method, So You Can Judge The Findings

We are documenting the method in full because an audit is only as trustworthy as the process behind it, and because we want you to be able to run a lighter version of this yourself.

Sample. Fifty Indian D2C brands, spread across five categories we see a lot of: beauty and personal care, food and beverage, apparel and accessories, health and wellness, and home and living. We chose brands with a real web presence and active marketing - not dormant Shopify stores - so that invisibility could not be blamed on the brand simply not existing online. These were brands trying to be found.

Question batteries. For each brand we built twelve to fifteen category questions. Crucially, these were not brand-name questions. Asking an assistant "tell me about Brand X" and getting an answer proves nothing - the engine can read the brand's own homepage. The real test is the unbranded category question, the one a shopper asks when they do not yet know the brand exists. We drew the questions from People Also Ask data, assistant autosuggest, and the genuinely messy phrasing people use on Reddit and Quora, so the battery reflected real demand rather than our guesses.

Engines. Every question was run across ChatGPT (with browsing), Perplexity, and Gemini. We started with Perplexity as the diagnostic anchor because it exposes its cited sources explicitly, which tells you not just whether a brand appeared but which page - the brand's own, or a third-party page about it - the engine actually leaned on. The tactical playbooks behind each engine, if you want them, are in How to Rank on ChatGPT and How to Rank on Perplexity.

Scoring. For every response we logged four flags: was the brand named, was it cited as a linked source, was the description accurate, and did competitors appear in its place. The headline metric - the visibility score - is simply the share of relevant prompts across all three engines where the brand was named or cited. A brand that surfaced in nine of forty-five brand-battery responses scored twenty percent. Simple, and brutally clear.

The Findings: Most Of A Category Is Missing

Here is the distribution, which is the whole story in one table.

Three observations sit underneath those numbers, and each one matters more than the headline.

Google rank did not predict AI visibility. This was the strongest and most uncomfortable finding. We cross-checked each brand's classic Google visibility for its head terms against its AI score, and the correlation was close to nothing. Brands ranking on page one for their money keywords sat at zero AI visibility. A couple of brands with mediocre Google positions scored better in AI answers than their higher-ranking rivals, because they happened to have the things AI engines reward. If you take one thing from this study, take this: your Google dashboard is telling you nothing about whether an assistant will recommend you.

Invisibility clustered by behaviour, not by category. It was tempting to expect one category to be structurally harder. It was not. Every category had invisible brands and every category had at least one that surfaced. What separated them was not what they sold but how they had built their web presence. That is good news, because it means visibility is a function of choices you control, not of the market you happen to be in.

Accuracy was a second, hidden failure. Among the eleven brands that did surface at all, the description was wrong or outdated in a meaningful share of appearances - a brand described by an old positioning it had abandoned, or conflated with a similarly named competitor. Being named is necessary but not sufficient. Being named correctly is the actual goal, and it is a separate problem with its own fix, which we get into in AI Citation Tracking.

The Five Reasons Brands Were Invisible

When we traced each low score back to its cause, the same five failures kept recurring. Almost every invisible brand had at least three of them. They are listed in rough order of how often they were the decisive factor.

1. Content written for keywords, not questions

The most common failure by a distance. The invisible brands had content - often a lot of it - but it was built for the old game. Pages engineered around a keyword, with the actual answer to a buyer's question buried under three hundred words of preamble, or never stated plainly at all. AI engines retrieve self-contained passages that answer a specific question. A page optimised for "best face serum" that never cleanly answers "which face serum should I use for pigmentation on Indian skin" gives the engine nothing to extract. The fix is a structural rewrite toward question-shaped, extractable content, which is the core of Answer Engine Optimization and the distinction we draw out in SEO vs AEO vs GEO.

2. No editorial content at all - only product pages

A large share of invisible brands had no explanatory content whatsoever. Their entire web presence was a homepage, a set of product pages, and a checkout. There was nothing for an engine to cite when answering an informational question, because the brand had never published anything that answered one. When a shopper asks how to choose a product in the category, an engine reaches for pages that explain how to choose - and a pure-commerce site has none. This is the gap content marketing exists to close, and in an AI-first world it is no longer optional decoration; it is the raw material engines quote from.

3. No third-party corroboration

AI engines are cautious about taking a brand's word for itself. They weight sources they can verify independently - reviews, Reddit and Quora discussions, press coverage, comparison articles, listicles from publications. The invisible brands were almost entirely un-discussed by anyone but themselves. No review threads, no earned mentions, no third-party comparisons. An engine assembling a recommendation had no corroborating source to lean on, so it reached for brands that other people, not just the brand, talked about. Earned mentions are slow to build, but they are decisive, and they are why we increasingly treat digital PR and topical authority as an AI-visibility play rather than only a link-building one.

4. Crawlability and rendering barriers

A quieter but fatal category. Several invisible brands were technically unreachable. Some had storefronts where the meaningful content only appeared after JavaScript execution that crawlers did not run, so an engine saw an almost empty page. A few had inadvertently blocked AI crawlers, or buried their content behind interactions no bot would trigger. You cannot be cited from content an engine never sees. This is unglamorous technical SEO work, but for these brands it was the entire ballgame - everything upstream was moot until the content was renderable and reachable.

5. No machine-readable entity

The subtlest failure, and the one that separates the merely-present from the reliably-recommended. The visible brands existed as clear entities - a consistent name, structured data describing what they are, and often a knowledge-panel or Wikidata presence tying the brand to its category. The invisible brands were, to a machine, just a string of characters that appeared on a website. Nothing told the engine "this is a D2C sunscreen brand operating in India." Without that entity anchor, even good content struggles to be connected to the category questions it should answer. Defining the entity cleanly is foundational, and it is one of the first things we assess in an AI SEO engagement.

What The Three Visible Brands Did Differently

The three reliable brands were not the biggest in the sample, and two of them were not the best-funded. They were simply the ones that had, deliberately or by luck, done the opposite of the five failures.

All three had real editorial content answering category questions in clean, extractable passages - buying guides, ingredient explainers, comparison pages that named the trade-offs honestly. All three had third-party corroboration: they were talked about on Reddit, reviewed by independent sites, mentioned in publication round-ups. All three had a clear entity - consistent naming, structured data, and a machine-legible tie between the brand and its category. All three were fully crawlable, with content that rendered without a browser. And all three had topical depth - enough coverage of their category that they read as an authority on it, not as a single product hoping to be found.

None of this was exotic. There was no growth hack. The visible brands had simply built a web presence that answered questions, that other people vouched for, that a machine could read, and that covered its category with enough depth to be believable. That is the entire recipe. It is boring, it compounds, and almost nobody in the sample had done it - which is precisely why the three that had were sitting alone in the answers.

The Framework: Is Your Brand AI-Recommendable?

We turned the five traits of the visible brands into a working scorecard, because a diagnosis you cannot score is just an opinion. This is the same framework logic behind how we decide which clients are worth a GEO strategy - if you are honest about where you sit on these five, you will know exactly what to fix first.

Score your brand from 0 to 2 on each dimension - 0 for absent, 1 for partial, 2 for strong.

• Editorial content. Do you publish content that answers category buyer questions in clean, extractable passages? Or is your site only product pages and a checkout?
• Third-party corroboration. Does anyone other than you talk about your brand - reviews, Reddit, Quora, press, comparison articles? Or are you the only source that mentions you?
• Machine-readable entity. Is your brand a defined entity - consistent name, structured data, ideally a knowledge-panel or Wikidata presence tying you to your category? Or are you just a string on a website?
• Crawlability and rendering. Can AI crawlers reach and render your meaningful content without executing a browser? Or is it hidden behind JavaScript, interactions, or a bot block?
• Topical depth. Do you cover your category with enough breadth to read as an authority? Or do you look like a single product hoping to be found?

Add it up. A brand scoring 8 to 10 is genuinely AI-recommendable and, in most categories, will be one of the few names that surfaces. A brand at 4 to 7 is inconsistently visible - present on one engine, absent on others, often described inaccurately. A brand at 0 to 3 is invisible, and no amount of Google ranking will change that. Seventy-six percent of the brands we audited scored 3 or below.

What To Do With This, If You Scored Low

If you ran your brand through that scorecard and did not like the number, the sequence to fix it is not mysterious, and the order matters.

Start with crawlability, because everything else is moot if engines cannot see your content. Then build the editorial layer - question-shaped, extractable content answering your category's real buyer questions, which is both an AEO and a content discipline. In parallel, define your entity cleanly with structured data and consistent naming. Then, over months, earn the third-party corroboration that turns a present brand into a recommended one. And measure the whole thing with a repeating audit, because AI visibility is a trend to be moved, not a box to be ticked. If you would rather have someone run the full audit and the rebuild for you, that is precisely what our AI SEO services and SEO audit work exist to do, and it is the fastest-moving part of what makes us, on the evidence, one of the SEO agencies in India taking this shift seriously rather than talking around it.

The larger point is the one the numbers make on their own. Seventy-six percent invisible is not a crisis to be feared; for a brand willing to do the work, it is the cheapest share-of-voice opportunity in a decade. When almost an entire category is missing from the answers, the brand that shows up owns the recommendation. That window is open right now, in most D2C categories, because the work is unglamorous and hardly anyone has started. It will not stay open. The brands that move first - the D2C brands treating this as a core channel now rather than a curiosity to revisit later - are building a lead that late movers will spend years trying to close.

We ran this audit expecting to confirm a hunch. We came away with something more useful: proof that the game is wide open, and a scorecard for winning it.

Frequently Asked Questions

What is an AI visibility audit?

An AI visibility audit measures how often, and how accurately, an AI assistant names or cites your brand when someone asks the category questions your buyers actually ask. You assemble the real buyer questions for your category, run each across ChatGPT, Perplexity and Gemini, and record whether the brand was named, cited, described accurately, or absent. The output is a visibility score plus a diagnosis of why you did or did not appear - the closest thing there is to seeing your brand the way an assistant sees it.

Why was my brand invisible in AI answers even though it ranks on Google?

Because AI assistants and Google reward different things. A page can rank by matching a keyword and earning links, yet give an AI engine nothing to extract - no clean answer, no corroborating source, no machine-readable entity. In our audit, Google ranking had almost no correlation with AI visibility. Treating the two as the same channel is the most expensive assumption a D2C brand can make right now.

How do I measure whether my brand shows up in ChatGPT or Perplexity?

Test it directly. Build twelve to fifteen unbranded category questions, run them across the assistants, and log for each response whether you were named, cited, described accurately, or replaced by a competitor. Start with Perplexity because it shows its cited sources, which tells you which of your pages the engine actually trusted. Repeat monthly to turn the snapshot into a trend.

What makes a brand AI-recommendable?

Five things, in our study: genuine editorial content that answers category questions, third-party corroboration the engine can verify, a clear machine-readable entity, full crawlability, and topical depth. A brand strong on all five is recommendable; most brands we audited were strong on none.

How long until my brand becomes visible after fixing these issues?

Longer than classic SEO and less linear. Live-retrieval engines like Perplexity can move within one to three months of a serious rebuild once your content is reachable and question-shaped. The corroboration and entity layers compound over months, and training-data-heavy engines update on their own slower cycle. It is a compounding asset, not a switch.