The Citation Gap: We Tracked 1,200 Brands Inside ChatGPT for 90 Days

From January through March 2026 we tracked citation behaviour for 1,200 DTC brands across ChatGPT, Claude and Perplexity. We ran the same 90 category-defining queries each week, recorded which brands were named in responses, which URLs were cited, and which queries triggered the highest citation count. The dataset is now 36,000 query-responses with full citation logs.

The headline finding: citation distribution is sharply non-uniform. Roughly 6% of tracked brands account for 73% of total citations. That ratio is more extreme than search ranking distributions in 2018, and it is widening. This piece walks through the findings, the structural drivers, and what brands outside the top 6% should do.

Method

We sampled brands across 12 categories — apparel, beauty, food and beverage, home goods, fitness, baby, pet, jewellery, electronics, footwear, eyewear, and outdoor. For each category we curated five "discovery" queries ("what are the best..."), three "constraint" queries ("...under £100"), and two "direct" queries ("tell me about brand X"). Total 120 queries per category times three engines, run weekly for 13 weeks.

We logged every brand named in responses and every URL cited. We cross-referenced brands against our own platform data to control for review volume, llms.txt presence, and structured-data completeness.

Top-line numbers

• 1,200 brands tracked; 947 received at least one citation across the 13 weeks.
• 73% of total citations accrued to 71 brands (5.9% of the sample).
• Top brand averaged 412 citations per week; bottom-of-cited group averaged under 4.
• 253 brands received zero citations despite being in established categories with healthy organic traffic.
• Citation count correlated 0.84 with verified-buyer review volume on the brand's PDPs.

What distinguishes the top-cited brands

We ran a structural audit on every brand that received 50+ citations per week. Five attributes are nearly universal in the top tier:

1. Clean Product + Offer + AggregateRating + Review schema on every PDP.
2. A populated FAQPage on each top SKU with 5+ questions per page.
3. An llms.txt file with the brand grounding paragraph and 12-30 priority pages.
4. Open robots.txt for GPTBot, ClaudeBot, PerplexityBot, Google-Extended.
5. A reviews corpus of 200+ verified-buyer reviews per top SKU.

The 253 zero-citation brands had on average two of those five attributes. The top-cited tier averaged 4.8 of them.

Engine-level divergence

ChatGPT, Claude and Perplexity do not cite the same brands.

• ChatGPT: heaviest weighting on AggregateRating and brand-mention frequency in third-party press. Wikipedia presence is a strong amplifier.
• Claude: weights FAQ and resource-hub content more heavily. Brands with strong /resources surfaces over-index on Claude citations.
• Perplexity: weights Reddit and forum mentions, then product-spec attribute density. The most evidence-greedy of the three.

A brand optimising for any single engine misses 40-60% of available citations across the trio. The right target is the union.

Category-level effects

Some categories show wider citation gaps than others.

• Beauty and electronics had the widest dispersion — top brand outcited bottom-cited by 100x or more.
• Outdoor and apparel were more even — top brand outcited bottom by ~20x.
• Pet, baby, and food had the thickest "middle" — many brands receiving moderate citations rather than a few dominant ones.

The narrower the dispersion, the more accessible the citation game is for a new entrant. The wider the dispersion, the more the top tier is structurally entrenched.

What changed over 90 days

Citation distribution is not static. Two patterns over the study window:

1. The top 71 brands gained share on average; their combined citation share rose from 71% to 75% over 13 weeks. Compounding visibility.
2. A small cohort of 30 brands broke into the top tier from the middle. The common move: shipping llms.txt and aggressively adding FAQPage schema in the first month.

Citation share is gainable, but the marginal cost of breaking in is rising. The brands moving up moved fast in the first six weeks of the year; later entrants in the study window saw smaller lifts.

What to do

If you are outside the top tier in your category:

1. Audit your current state against the five universal attributes. Most brands have 2 of 5.
2. Ship the missing three within 60 days.
3. Set up a weekly citation tracker. At minimum, run the top 10 queries in your category each Monday and log who is cited.
4. Pick the engine where you under-perform most relative to your share of voice elsewhere. Diagnose what is missing.
5. Treat citation share as a north-star KPI alongside paid CAC and organic sessions.

Forecasts

Three predictions for the next 18 months based on the data.

• Citation share concentration will continue to widen. The top 6% will become the top 4% by year-end.
• Engine-specific optimisation will become a meaningful sub-discipline. Expect AEO consultancies to start splitting into "ChatGPT-led", "Claude-led" and "Perplexity-led" specialisations.
• Brands without a citation strategy by Q4 will see organic acquisition costs rise 15-30% as AI-engine referrals continue to grow as a share of total organic traffic.

Closing

Citation distribution looks like 2008 PageRank — extreme concentration, hard to crack from the bottom, lucrative at the top. Unlike PageRank, the structural fixes are well-bounded and ship in weeks. The window for breaking into the top tier is open but narrowing.