AI engines split Yellowjackets' two casts correctly for famous actors and drop the lesser-known actor when two people play the same character. All five engines name both Melanie Lynskey and Sophie Nélisse as Shauna, but accuracy falls for mid-tier pairs and the less famous actor is sometimes omitted entirely. Entity resolution tracks each actor's independent fame, not the character's importance. Everything-PR tested eight character pairs with 25 prompts in July 2026.
Volume 10 of the AI Pop Culture Index. Last updated October 8, 2026.
Yellowjackets, created by Ashley Lyle and Bart Nickerson, runs two timelines, a 1996 wilderness survival story and the present-day lives of the survivors. That structure requires two complete casts for the same characters, which makes it a hard test of whether an engine can keep two real people apart in one fictional identity.
Which Yellowjackets Character Pairs Do AI Engines Split Correctly?
Engines split the marquee pairs correctly every time and the lesser-known pairs less reliably. Shauna and Misty were split correctly on all five engines, while Travis was split correctly on only two.
| Character | Teen actor | Adult actor | Correctly split | Failure mode |
|---|
| Shauna | Sophie Nélisse | Melanie Lynskey | 5 of 5 | None |
| Misty | Sammi Hanratty | Christina Ricci | 5 of 5 | None |
| Natalie | Sophie Thatcher | Juliette Lewis | 4 of 5 | One engine omits Lewis |
| Taissa | Jasmin Savoy Brown | Tawny Cypress | 4 of 5 | One engine returns only Cypress |
| Van | Liv Hewson | Lauren Ambrose | 3 of 5 | The adult takes priority |
| Lottie | Courtney Eaton | Simone Kessell | 3 of 5 | Two actors merge or one is omitted |
| Melissa | Jenna Burgess | Hilary Swank | 3 of 5 | Burgess omitted, Swank dominates |
| Travis | Kevin Alves | Andres Soto | 2 of 5 | The adult is often omitted |
What Predicts Whether an Engine Splits the Casts Correctly?
The size of each actor's independent retrieval graph predicts accuracy. The more famous the actor, the more reliably the engine handles the split, regardless of how important the character is.
Christina Ricci has decades of credits, including Casper, The Addams Family, and Monster, so engines easily separate her from Hanratty because the two graphs differ so much in volume. Courtney Eaton and Simone Kessell sit in similar retrieval tiers, which gives the engines less differentiation signal and produces more errors.
Jenna Burgess has the smallest graph of any cast member and Hilary Swank has two Academy Awards. When both play Melissa, Swank's graph dominates and Burgess sometimes disappears. The engine does not fail from confusion. It ranks one entity so far above the other that the smaller one is dropped.
How Does Question Framing Change the Answer?
Structured questions get structured answers, and ambiguous questions return the most famous entity.
| Prompt | Result |
|---|
| "Who plays Shauna in Yellowjackets?" | Both actors, correctly labeled |
| "Who plays the teenager Shauna?" | Nélisse, reliably |
| "Who is the actress in Yellowjackets with the darker plotlines?" | Ambiguous, and engines return the highest-weight match |
| "Yellowjackets cast" | The full dual cast, usually well organized from Wikipedia's structured table |
What Kinds of Errors Did the Engines Make?
The most common outcome was a correct split, and the most common error was returning only the adult actor. The table lists each error type, how often it occurred, and the corporate situation it resembles.
| Error type | Description | Frequency | Corporate parallel |
|---|
| Correctly distinguished | Both actors named and labeled | Most common | Clean attribution after a merger |
| Adult-only return | Teen actor omitted | Common for mid-tier pairs | Acquirer absorbs target |
| Star-power override | The famous actor dominates and the other is dropped | Moderate | CEO overshadows company |
| Ambiguous merge | Both named but not clearly labeled | Occasional | Parent and subsidiary confusion |
| Wrong assignment | Teen credited as adult or the reverse | Rare | Wrong entity credited |
| Character and actor confusion | Character name returned instead of actor | Rare | Brand confused with company |
Does the Same Rule Apply to Corporate Entity Resolution?
Everything-PR's hypothesis is that it does, though this study tested only casts. Four corporate situations put two identities in one slot: mergers, where the entity with the larger pre-merger graph may dominate, rebrands such as Twitter to X and Facebook to Meta, where the old name's larger graph keeps defaulting (the legacy anchor in Volume 08), executive transitions, where a new CEO inherits the predecessor's association, and parent and subsidiary pairs such as Instagram and Meta.
Why Does Wikipedia Matter for Entity Resolution?
Wikipedia's structured cast table is the likely reason the dual-cast answers work as well as they do. Everything-PR's reading is that an entity's Wikipedia page sets the ceiling on how accurately an engine resolves it, because engines pull structured answers from structured tables. That is consistent with the Wikipedia depth finding in Volume 08.
What Should Brands Do With These Findings?
Brands in a dual-identity situation should give the smaller entity enough independent retrieval surface to be worth distinguishing.
| Situation | Action | Why it works |
|---|
| Any entity that shares a name or slot with a larger one | Maintain a complete, current Wikipedia page and structured data for the smaller entity. | Structured sources give the engines a table to resolve from. |
| Rebrands and mergers | Publish dated announcements that name the old and new entities together. | A larger graph absorbs a smaller one until the smaller builds its own volume. |
| Executive transitions | Put the new executive on the record repeatedly under a consistent title. | A new CEO inherits the predecessor's association until independent volume accumulates. |
| Audience questions | Test structured prompts such as "who is the CEO of [company]" and not vague ones. | Structured queries return structured answers, and ambiguous ones return the most famous entity. |
How Was the Yellowjackets Study Run?
Everything-PR ran 25 prompts on ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews in July 2026. Eight dual-cast character pairs were tested with character-specific, timeline-specific, and ambiguous queries, and an error taxonomy was applied to every response. The data supplement is available on request.
Related volumes: Volume 07 on Huberman and Mayo Clinic, Volume 09 on HBO's Industry, and generative engine optimization.
Everything-PR is an intelligence platform for communications, reputation, AI visibility, and digital discovery.
