¶ AI Citation Index — July 2026
Who does AI cite when America asks about real estate?
We ran 612 real homebuying and selling questions through five AI answer engines — 15,300 answer runs — and recorded every source they cited. This is what we found.
- 1Zillow14.7%
- 2Realtor.com11.4%
- 3Redfin9.5%
- 4Investopedia7.7%
- 5Reddit6.6%
n = 15,300 runs · ±1.2 pp · CC-BY 4.0
612
Prompts tested
Weighted by real search demand
15,300
Answer runs
5 runs per prompt per platform
5
AI platforms
ChatGPT, AIO, Perplexity, Copilot, Claude
41,286
Citations captured
±1.2 pp at 95% confidence
The ranking
July 2026 — top 10 of 1,912 cited domains
| # | Source | Type | Citation share | Δ Jun | 6-mo trend |
|---|---|---|---|---|---|
| 1 | Zillowzillow.com | Marketplace | 14.7% | -0.8 pp | |
| 2 | Realtor.comrealtor.com | Marketplace | 11.4% | -0.3 pp | |
| 3 | Redfinredfin.com | Brokerage | 9.5% | +0.4 pp | |
| 4 | Investopediainvestopedia.com | Publisher | 7.7% | +0.2 pp | |
| 5 | Redditreddit.com | Community | 6.6% | +1.1 pp | |
| 6 | Bankratebankrate.com | Publisher | 5.9% | -0.2 pp | |
| 7 | NARnar.realtor | Association | 4.5% | +0.1 pp | |
| 8 | NerdWalletnerdwallet.com | Publisher | 4.1% | +0.3 pp | |
| 9 | Homes.comhomes.com | Marketplace | 3.8% | +0.5 pp | |
| 10 | HUDhud.gov | Government | 3.5% | +0.2 pp |
What the data says
Key findings
- Finding 01
The portal triopoly is loosening
Zillow, Realtor.com and Redfin still absorb 35.6% of all citations, but their combined share is down 3.1 pp since February. The losses are not going to other portals — they are going to community discussions and video, which grew from 19.1% to 22.4% over the same period.
- Finding 02
Each AI engine has a distinct sourcing personality
ChatGPT over-indexes on reference content (Wikipedia 3.9%, Investopedia 8.9%). AI Overviews routes 3.2× more citations to community and video than ChatGPT. Claude is the most institutional, favoring government and association sources. A brand's AI visibility strategy cannot be platform-agnostic.
- Finding 03
Guides beat listings on informational intent
For how-to and definitional prompts — 58% of the corpus — explainer content captures 41% of citations while listing pages capture just 9%. Listing pages only win when intent is explicitly transactional. Content type, not domain authority, is the strongest predictor of citation on informational prompts.
Coverage
Industries under measurement
| Industry | Status | Prompt corpus | |
|---|---|---|---|
| US Real Estate | ● Live | 612 | View → |
| US Health Insurance | ◐ Collecting | 540 | |
| Consumer Banking | ○ Q4 2026 | — | |
| B2B SaaS | ○ Q4 2026 | — | |
| US Travel | ○ Q1 2027 | — |
One industry goes deep every month. New verticals are chosen by reader vote and data availability — the pipeline runs identically for each.
Radical methodology
Every number on this site comes from a published, repeatable protocol: a prompt corpus weighted by real search demand, five answer runs per prompt per platform, deterministic citation extraction, and monthly re-runs from a frozen prompt set. No screenshots, no anecdotes — and we publish our confidence intervals.
Read the full protocol →Underlying dataset
citations.csv41,286 rows
prompts.csv612 rows
domains.parquet1,912 domains
methodology.pdf14 pages
Free with email. CC-BY 4.0 — cite us, reuse freely.
Free visibility scan
Is AI citing your brand — or your competitor?
This index is a monthly snapshot. Pagelens tracks your citation share across every AI engine in real time, down to the prompt.