Trakkr Data

Citations

The sources AI answers cite most. Every URL the models lean on, aggregated across ChatGPT, Gemini, Perplexity and more, then ranked by domain with category, query intent, freshness and trend.

Citations tracked
20M
source links, across models
Sources cited
313K
unique domains
URLs cited
1.4M
distinct pages
Brands analysed
1.3K
tracked footprints

Citation activity

New source links logged per day across the brands we track.

Source: Trakkr Citation Index · per-day new citations · CC BY 4.0

The most-cited sources

Every domain AI answers reach for, ranked by total citations. Search across all sources or filter by type.

Where AI gets its answers

Named source types by share of citations, paired with how fresh each one's links are.

Plus 0.0% from brand-owned and niche domains — the long tail of 313K sources.

How long a citation lasts

Getting cited is only half of it. Here’s how fast a source link fades once it appears — and how quickly a brand’s whole footprint halves.

73.5%
are one-and-done

Most source links show up in a single answer, then never come back.

Mean lifespanhow long a citation lives6.8 days
Brand half-lifeto lose half your citations31 days
Halve in a weekof brands, within 7 days47%

Survival curve

Of every 100 citations, how many are still live N days later — only about 23 in 100 last two weeks.

20%40%60%80%100%
0d2d5d7d14d
Source: Trakkr · the half-life of AI citations · CC BY 4.0

The sources that stick

Domains whose citations last longest, by average lifespan in days.

1lodgify.com
32.2d
2hostaway.com
30.5d
3brusol.be
27.9d
4rippling.com
24.2d
5leadearly.co.uk
22.8d
6spothero.com
21.2d
7woocommerce.com
20.0d

Here today, gone tomorrow

High-volume domains whose citations rarely survive past the day they first appear.

1carfax.com
0.0d
2instagram.com
0.3d
3ebay.com
0.4d
4sciencedirect.com
0.4d
5autotrader.com
0.5d
6cars.com
0.6d
7kbb.com
0.8d

What people ask

The shape of the questions behind the citations, and where each kind pulls its sources from.

What earns a citation

Signals that move a page's odds of being the source AI quotes.

Title format, by citation lift

Multipliers are relative to the average page. Numbered listicles and the current year both signal the kind of fresh, scannable content AI prefers to quote.

What a cited page looks like
Methodology

Built from the source links AI returns across the brands Trakkr tracks, normalised to domains, categorised and recounted nightly. Shares are of all tracked citations. Freshness measures the portion of a type's citations that point to content under 60 days old.

Trakkr Citation Index·RankingsCC BY 4.0

Common questions

Which sources do AI engines cite most?

Across models, AI answers lean on a relatively small set of high-authority sources — major reference sites, established publishers and large community platforms — though the exact mix shifts by category and query intent. The Citation Index ranks every domain by citation count and updates daily.

Do different AI models cite the same sources?

There is a shared core of trusted domains, but each engine adds its own tilt. The dataset lets you compare a domain’s citation share and trend, and slice it by query intent, to see where models agree and diverge.

What counts as a citation in this dataset?

Every URL an AI engine references in an answer is recorded, normalised to its source domain, then aggregated across models and broken down by source category, query intent and freshness.