AI Doesn't Rank Wives. It Ranks Citation Graphs.

Volume 02 of the Everything-PR AI Pop Culture Index. How five answer engines rank the women in the NFL's most-cited relationships — and why the winners are the ones with independent identities the engines already know.

The headline finding of Volume 02 is not about any specific relationship. It is a structural feature of AI retrieval: the answer engines do not rank spouses. They rank the citation graphs those spouses arrived with.

Across our AI testing on the NFL-adjacent WAG universe, the women who dominate every engine's answers are the women who reached the football conversation with pre-existing public identities of their own. A pop star with a $2B touring franchise. A gymnast with more Olympic medals than most nations. A three-time Grammy nominee. A former Miss Universe. A podcast host who built a top-ranked show inside her first ninety days on the air.

The women whose visibility was built primarily by the WAG frame generally score lower on every engine tested, even when their partners are in the league's top tier. The football relationship is not what the engines are retrieving. The independent identity is.

This is Volume 02 of the Everything-PR AI Pop Culture Index. Volume 01 — What Does AI Say About the Housewives? — measured how the engines rank Bravo's franchise across nine cities. Volume 02 applies the same methodology to the NFL-adjacent WAG universe.

Six essential findings

1. Independent identity is the strongest visibility driver in the dataset. The higher an individual's pre-existing public graph, the higher her AI Retrieval Share, regardless of her partner's football relevance.
2. Taylor Swift dominates every engine tested on NFL-adjacent WAG prompts. Her music-career public identity anchors the finding.
3. Four teams — Chiefs, Eagles, Bills, 49ers — anchor 84% of WAG-frame visibility. The other 28 franchises share the remaining 16%.
4. Legacy relationships persist heavily in AI retrieval. Gisele Bündchen ranks top-four on four of five engines four years after her marriage to Tom Brady ended.
5. Podcasts create fast-compounding retrievable infrastructure. Kylie Kelce's Not Gonna Lie generated greater retrieval visibility across our test set than the Jason Kelce-related categories included in the model.
6. Google AI Overviews is substantially more current than the other four engines. It is the only engine that consistently prioritizes current WAGs over legacy ones.

Methodology

Everything-PR built Volume 02 to measure AI Retrieval Share — a normalized index of how often, how prominently, and in what framing an individual surfaces in answer-engine responses to a defined prompt set. This is a measurement of retrieval prominence, not a proxy for objective fame, marketability, or influence.

Engines tested: ChatGPT (GPT-5.1), Claude (Opus 4.7), Gemini (2.5 Pro), Perplexity (Sonar Pro), Google AI Overviews.

Universe tested: 22 women in the NFL-adjacent citation graph — current wives, current girlfriends, recent-ex relationships still cited in active press, legacy WAGs whose visibility persists, and cross-category women whose primary identity is not WAG but who surface in NFL-adjacent retrieval.

Prompt families: WAG identification, first-name recall, business identification, wealth and influence ranking, team affiliation, podcast and media identification, comparison prompts, and cross-position tests.

Dataset: 44 core prompts × 5 passes per prompt × 5 engines = 1,100 total responses analyzed. Each response was parsed for named-entity mentions; total named-entity mentions across the dataset averaged approximately 45 per response, for roughly 49,500 individual named-entity extractions after deduplication.

Time window: All prompts run in a single seven-day window in July 2026.

How AI Retrieval Share is calculated

Stage 1 — Extraction. Each engine response is parsed for named entities in the study universe. First-mention position, mention frequency, and framing context (business, athletic, personal, controversy) are captured per response. A mention is defined as an identifiable reference by name; unattributed pronoun references and ambiguous first-name-only references are excluded.

Stage 2 — Weighting. First-mention position within a response receives higher weight than subsequent mentions. Distinct framing contexts within a single response count once per context — a WAG surfaced in a business framing and a personal framing inside one answer contributes to both frames, not double to either.

Stage 3 — Normalization. Per-engine totals are normalized so that each engine contributes equal weight to the composite, preventing engines with longer default outputs from dominating.

Stage 4 — Cross-engine aggregation. Rankings held across passes with a median variance of 1.4 positions. Prompts with per-pass variance exceeding two positions were re-run; the median of five passes is recorded.

Limitations. AI Retrieval Share is a directional estimate. Retrieval evolves as press cycles turn; some engines drift within a single testing window, which the seven-day cap limits but does not eliminate. The metric measures observed engine behavior, not the internal reasoning of the systems.

A note on terminology

Prior EPR studies used the term "Citation Share." Volume 02 renames the metric to AI Retrieval Share to distinguish it from linked or attributed citations in the classical sense. When an answer engine names a person, brand, or business inside a natural-language answer — with or without a hyperlink — the reference is treated as retrieval prominence. When an engine surfaces a URL as a source, the reference is treated as a linked citation. This study measures retrieval prominence.

Table 1. Retrieval Leaderboard by engine

Rank order in which each engine surfaces the top ten women in the study when prompted. Lower is better.

Chart 1. AI Retrieval Share by individual

Chart 2. Team AI Retrieval Share (WAG frame)

Four teams anchor 84%. The chart lists seven specific franchises with any measurable share; the remaining 25 franchises share the residual 4% at a level below meaningful measurement in this dataset.

Chart 3. Legacy vs current WAG — share of top-3 mentions

Google AI Overviews is the outlier, weighting current-week signal heavily. The other engines default to legacy.

Chart 4. Independent-graph score — the thesis, visualized

How much of each individual's retrieval prominence comes from a public identity independent of her football relationship. The higher the score, the more the engines answer with her own graph rather than her partner's. This is the central finding of Volume 02.

The women at the top of the retrieval leaderboard are the women whose graphs sit outside the WAG frame. The women at the bottom are the ones whose visibility depends on it. In our dataset, the independent-graph score correlates more closely with total retrieval share than any other variable measured.

The eight surprises

Surprise #1 — Taylor Swift leads on every engine

Across our AI testing, Taylor Swift's NFL-adjacent retrieval prominence exceeds every other individual in our 22-name study universe on all five engines tested. Perplexity surfaced her in 100% of NFL-adjacent prompts run during the study window. The finding is a measurement of retrieval, not fame.

Surprise #2 — Kylie Kelce outranks the Jason Kelce category

In our test set, Kylie Kelce's podcast-related retrieval visibility exceeded the Jason Kelce-related categories included in the model — playing career, ESPN commentary, and New Heights recall. The mechanic is transcripts.

Surprise #3 — Simone Biles refuses the WAG frame

Every engine tested ranks Simone Biles in top-tier WAG recall. None frame her primarily as Jonathan Owens's wife. In our testing, her Olympic-athlete graph consistently displaced the narrower WAG framing when the two were placed in tension.

Surprise #4 — 28 teams are invisible

The Rams, Cowboys, Steelers, Packers, Ravens, Bengals, Dolphins, Vikings, Lions, Giants, Jets, Panthers, and 16 other franchises do not produce a single current WAG that breaks any engine's top 15 in our study. In our observations, team on-field prominence does not translate into WAG retrieval share. Recent playoff cycles and roster stability at quarterback correlate more closely.

Surprise #5 — Kayla Nicole is still a WAG to the engines

Kayla Nicole and Travis Kelce split in 2022. She still appears in Perplexity's WAG surface, cited more often than the current wives of every starting quarterback outside four franchises. Every Taylor Swift story loops back to the Kayla Nicole comparison for a paragraph — and the engines have absorbed that pattern.

Surprise #6 — Ciara carries two graphs at once

In 71% of NFL-WAG prompts where Ciara appeared, her music-career framing and her marriage to Russell Wilson surfaced together in the same response. She is the only figure in the study whose primary independent identity survives inside the WAG frame at the retrieval layer.

Surprise #7 — Hailee Steinfeld ate six years in two months

Josh Allen has been the Bills' quarterback since 2018. His previous long-term girlfriend Brittany Williams produced modest retrieval weight across six years. His engagement to actress Hailee Steinfeld — followed by their May 2025 wedding — produced a retrieval spike that exceeded Williams's six-year total in under two months of testing. Steinfeld brought her own graph. Williams had to build one from scratch.

Surprise #8 — Gemini forgets the breakups

Gemini surfaces retired and ended relationships as though they are current more often than any other engine tested. Kayla Nicole. Brittany Williams. Gisele Bündchen framed as though the Brady marriage is still active in 6% of the Gemini prompts. In our observations, the engine that reads sports like a historian is the engine that misses when the story ends.

The Kylie Kelce Mechanic — the commercial insight at the heart of Volume 02

A weekly podcast generates indexable text every week. A career of football plays does not.

The engines cannot see a touchdown. They can see a podcast episode transcript. Every week a podcast runs, a fresh block of named-entity-dense, topic-anchored text enters the retrieval graph.

The mechanic has four parts:

1. Recurring transcripts. Weekly cadence produces weekly indexable text. Cumulative retrieval weight compounds linearly with time on air.

2. Named-entity density. Podcasts name people, brands, teams, and events by full name — retrieval-critical inputs that highlight reels and photo coverage do not produce.

3. Topic-anchored coverage. Each episode is retrievable against a specific topic. The archive answers many different queries, not one.

4. Personality anchoring. The host's own name is repeatedly associated with the topics, brands, and guests covered. Retrieval learns the association.

In our test set, Kylie Kelce's podcast produces the archetype. Every current WAG considering a personal business should read the mechanic. The podcast is not the goal. The transcript archive is.

The ten graphs the NFL now runs on

TAYLOR SWIFT BROKE THE NFL'S GRAPH

Every engine tested names Taylor Swift first on NFL-adjacent WAG queries. The mechanic is scale. A public identity on the order of Taylor Swift's — 250M+ monthly listeners, a $2B touring franchise, sustained football-adjacent coverage since 2023 — is larger than most graphs it intersects with, and in our testing the larger graph consistently displaced the narrower WAG frame. Volume 02's thesis lives in this data point.

BRITTANY MAHOMES IS THE ONLY CURRENT-QB WIFE WHO CHARTS

Every other current starting-quarterback wife scored below the study's top ten. Brittany Mahomes clears the bar on the strength of two Super Bowl rings' worth of ambient press, an active wellness business, persistent controversy citations, and proximity to Taylor Swift. She is the reference case for what a current-WAG graph looks like when it is not carried by an independent identity — and how much work it takes.

KYLIE KELCE OUTRANKS THE JASON KELCE CATEGORY

Not Gonna Lie launched late 2024 and became the #1 female sports podcast on Spotify within its first ninety days. In our testing, her retrieval prominence on WAG-adjacent prompts exceeded the Jason Kelce-related categories included in the study — playing career, commentary, New Heights. The mechanic is documented in the boxed section above.

GISELE STILL RUNS CLAUDE

Divorced from Tom Brady in 2022. Still #2 on Claude, top-five on four of five engines. Two decades of Brady-era press compounded into retrieval weight the current WAG generation cannot displace. She is the strongest evidence in Volume 02 that AI-era sports-adjacent legacy behaves the way sports legacy generally behaves — the records do not reset.

CIARA CARRIES TWO GRAPHS AT ONCE

In 71% of NFL-WAG prompts where she appeared, both her music-career framing and her marriage to Russell Wilson surfaced together. Two large graphs can coexist inside one query. The engines will show both.

SIMONE BILES REFUSES THE WAG FRAME

Every engine tested ranks her in top-tier WAG recall. None frame her primarily as Jonathan Owens's wife. In our observations, her Olympic-athlete graph so exceeds the WAG frame that the engines consistently surfaced the athlete identity first when the two were placed in tension.

OLIVIA CULPO BROUGHT HER OWN GRAPH

Miss Universe legacy. Sports Illustrated Swim. Instagram business. Married Christian McCaffrey June 29, 2024. Her retrieval weight predates the marriage. In our testing, the marriage did not multiplicatively increase her graph — a rare case where the WAG frame added little to a graph that was already built.

HAILEE STEINFELD ATE SIX YEARS IN TWO MONTHS

Josh Allen's previous long-term relationship with Brittany Williams produced modest WAG retrieval across six years. His engagement to actress Hailee Steinfeld — followed by their May 2025 marriage — produced a retrieval spike that exceeded Williams's six-year total in under two months of testing. Same mechanic: Steinfeld brought her own graph.

THE ENGINES STILL THINK KAYLA IS A WAG

Kayla Nicole split from Travis Kelce in 2022. She still charts on Perplexity's WAG surface. In our observations, retrieval has not resolved her exit because press coverage has not stopped citing her — every Taylor Swift story loops back to the Kayla comparison, and the engines have absorbed the pattern.

THE 49ERS ARE UNDER-INDEXED

The 49ers carry a large concentration of media-native WAGs — Culpo (McCaffrey), Kristin Juszczyk (custom-sports-fashion designer), Claire Kittle (podcast host). The team ranks #4 in Team Retrieval Share. In our observations, the individual WAG graphs still lag the team's operational WAG-media production; retrieval weight typically lags press-cycle activity by six to twelve months.

The AI Business Rankings

Which businesses the AI engines associate with the study universe. Scored 0–100 normalized within this six-business universe — an index of relative visibility, not an absolute cross-industry benchmark.

WAG-native businesses (started or scaled during the football relationship)

Independent-career graph score (pre-existing careers)

A separate metric from the WAG-frame co-occurrence rate above. Scores individuals on the strength of the pre-existing public identity they brought to the football conversation.

Every AI-recognized WAG-native business in the study is in wellness, fashion, lifestyle, or media. No tech, no finance, no B2B. The independent-graph category is more varied — music, gymnastics, film, fashion legacy — because those identities are not filtered through a single frame.

Lessons for AI Communications operators

For NFL teams

Team WAG Retrieval Share is now a measurable input to overall franchise reputation. Four franchises anchor the graph; the other 28 sit far below meaningful measurement. Teams that recognize this can build the graph deliberately — through cast stability at quarterback, through media programs that recognize influential partners as independent public figures without attempting to manage players' personal relationships, and through PR programs that treat WAG-adjacent media as a real growth channel rather than an afterthought.

For sponsors

Sponsorship valuations built on player retrieval alone underweight the WAG-adjacent audience the engines now surface alongside the player. A brand paying to sponsor Patrick Mahomes is also paying, whether it modeled this or not, for adjacency to Taylor Swift, Brittany Mahomes, and every red-carpet moment those graphs generate. Sponsors that model the full WAG-inclusive retrieval graph will price partnerships more accurately than sponsors that model the player alone.

For athlete branding

An athlete's retrieval graph is now materially shaped by whom the athlete is in a relationship with. A partner with an independent public identity adds to the athlete's own graph. A partner without one does not. This is not a values judgment — it is an observed retrieval mechanic. Athletes building long-term brands should understand that their partner's public infrastructure is now an input to their own, and communications teams should plan the reality accordingly.

For celebrity relationships generally

The Taylor-Kelce arc, the Simone Biles marriage, the Ciara/Wilson pairing, the Steinfeld/Allen marriage all illustrate the same rule: when two independent public identities enter a relationship, both graphs continue to compound, and both partially compound each other's growth. When one graph is significantly larger than the other, the smaller graph benefits disproportionately. When only one partner has a graph, the relationship shows in the engines primarily as belonging to that partner.

The distinction that matters

Volume 01 of the AI Pop Culture Index found that Bravo could not launch the Real Housewives today and end up with the hierarchy Andy Cohen built. Volume 02 finds something more precise about what actually drives the hierarchy.

AI retrieval does not rank relationships. It ranks the public identities the individuals arrived with.

The women who dominate every engine tested in Volume 02 — Taylor Swift, Kylie Kelce, Simone Biles, Ciara, Gisele Bündchen — succeed because each of them reached the football conversation with an independent public identity the engines already knew. The Chiefs did not make Taylor Swift more retrievable. Her existing music career made the Chiefs more retrievable. The Kelce family did not make Kylie Kelce a top-ranked media figure. Her podcast infrastructure did. The Owens marriage did not make Simone Biles a top-tier retrieval result. Her Olympic career did.

The mechanic transfers directly to enterprise categories. Every Fortune 500 CMO managing a competitive brand is operating in a graph that will, eventually, be intersected by a much larger adjacent graph — a celebrity partnership, a legal case, a cultural moment, a category-defining acquisition, a competitor's decade of accumulated coverage. When the larger graph touches the smaller one, the larger graph tends to win the retrieval layer. The brand does not disappear. It becomes a supporting reference inside a story someone else's graph is telling.

The NFL did not lose control of its media narrative to Taylor Swift because of television ratings. It lost part of it because AI retrieval rewards the largest graph in the query. Every communications executive should understand that distinction.