Spirit Halloween has one job. For about ten weeks a year it takes over empty storefronts and sells nothing but Halloween. If anyone should own the answer to "where do I buy a Halloween costume?", it's them.
So we asked the robots. Butter Labs ran 34 Halloween shopping questions through ChatGPT, Perplexity, Google AI Overviews and Copilot on October 24, 2026. The questions ranged from costumes to candy to the famous 12-foot skeleton. Then we counted which brands each answer named and which ones it linked to.
The brand AI named most often was not Spirit Halloween. It was Walmart, in 42% of answers. Spirit Halloween came in at 41%, with Amazon (37%) and Target (34%) close behind.
A one-point gap is not a landslide, and we won't pretend it is. The point is that the specialist, the brand that exists only for this one holiday, is fighting for first place against stores that also sell tires and toothpaste. If you run a seasonal or specialist business, that should make you a little nervous.
The AI Halloween scoreboard
Here is how often the most-mentioned brands showed up, and how often AI actually linked to their own website:
- Walmart: named in 42% of answers, linked in 25%
- Spirit Halloween: named in 41%, linked in 20%
- Amazon: named in 37%, linked in 19%
- Target: named in 34%, linked in 24%
- HalloweenCostumes.com: named in 19%, linked in 8.1%
- Party City: named in 16%, linked in 8.1%
- Home Depot: named in 14%, linked in 5.9%
Two things jump out. The top four are close together, and everyone gets named a lot more often than they get linked. Being in the answer and being the place AI sends the shopper are two different wins.
Each AI has a different favorite store
The four engines did not agree on much. Each one had a clear favorite, and they were all different:
- ChatGPT leaned Walmart (named in 59% of its answers)
- Perplexity leaned Target (38%)
- Copilot leaned Amazon (65%)
- Google AI Overviews leaned Spirit Halloween (54%)
Only Google put the Halloween specialist on top. On the other three engines, a general retailer won.
This matters because "how do we show up in AI?" is really four questions, not one. Each engine pulls from different sources and weighs them differently. Doing well on Google's AI Overviews tells you very little about how you look in ChatGPT, and the reverse is true too. If you only check one, you are guessing about the other three.
Why the big-box stores keep winning
Our data shows who won, not exactly why. But looking at the answers, a few patterns stand out, and they line up with how these tools generally build an answer.
They have a page for everything. Walmart, Target and Amazon have a product page for nearly every costume, inflatable and bag of candy. When an AI tool goes looking for a page to back up its answer, there is almost always one there to cite.
They show up across every aisle. A specialist is strong in one category. A big-box store is decent in all of them. Across 34 different questions, being decent everywhere adds up to more mentions than being great in one place.
Other sites talk about them. Deal sites, gift guides and forum threads mention the big retailers constantly. AI tools lean on those third-party sources, so brands that get talked about a lot get repeated a lot.
None of that requires a giant budget to copy. It requires clear pages and being talked about in the right places. More on that below.
Being named is not the same as being linked
Spirit Halloween was named in 41% of answers, but its own website was cited in only 20%. That is roughly half. Walmart's ratio was better (42% named, 25% linked), and Target's was the best of the top four (34% named, 24% linked).
A mention builds awareness. A link sends the shopper somewhere they can actually buy. If AI says your name and then links to a competitor, a marketplace or a coupon site, you did the work and someone else gets the click.
So where were the links going instead? These were the outside sites AI cited most:
- reddit.com: 9.6% of answers
- shopback.com: 8.8%
- youtube.com: 5.1%
- thehorrordome.com: 3.7%
Reddit and YouTube are where AI goes to check what real people think. A cashback and deals site made the list. And a niche horror site, thehorrordome.com, made the top four, cited in 3.7% of answers alongside Reddit and YouTube. That's the hopeful part for smaller brands: a focused site that covers its topic well can earn a spot in the answers. Learn more about how often AI cites you in our glossary entry on citation rate.
Spirit Halloween disappears outside the costume aisle
We grouped the questions by what people were shopping for. Costumes are Spirit Halloween's home turf, and it wins there. Step into any other aisle and it fades fast:
- Costumes and accessories (11 prompts): Spirit Halloween led with 24 answers, ahead of Amazon (20).
- Decorations (7 prompts): Home Depot led with 14. Spirit Halloween tied for second with Walmart (8 each).
- Party supplies and extras (7 prompts): Walmart led with 14, then Target with 12. Spirit Halloween got 6.
- Candy (6 prompts): Hershey led with 8, then Walmart with 7. Spirit Halloween got 0.
- Best Halloween store overall (2 prompts): Spirit Halloween led with 8, ahead of Amazon and HalloweenCostumes.com (5 each).
- Pet costumes (1 prompt): Petco and Target tied with 2 each. Chewy and PetSmart got 1 each. Spirit Halloween got 1.
In 12 of the 34 prompts, no AI named Spirit Halloween at all.
Notice that Spirit does win when someone asks for "the best Halloween store." AI clearly knows what Spirit Halloween is. It just doesn't connect that brand to decorations, party supplies or candy, even though Spirit sells plenty of decorations.
That is the real lesson. AI doesn't think of a brand as "the Halloween store." It thinks product by product and question by question. Whoever has the clearest, most cited page for that specific product gets the answer. Your brand reputation doesn't automatically carry over from one category to the next.
AI still sends people to stores that closed
Some of the strangest answers involved stores that are no longer there.
Joann closed all of its stores by May 2025. A Google AI Overview still told shoppers to try "Michaels & Joann Fabrics" for Halloween craft supplies. Party City closed its own stores in early 2025 too, and AI still named it in 16% of answers. In fairness, most of those answers pointed to Party City's relaunched website, which is open for business.
AI tools don't always notice when the world changes. Some of what they "know" comes from older information, and old blog posts, directories and listings stick around online long after a store closes. (Our glossary explains the idea of a knowledge cutoff.)
If your business moved, closed a location, changed its hours or rebranded in the last couple of years, there's a real chance some AI answers still describe the old version of you.
The scariest part: wrong hours and old prices
Shopping answers were only half the test. We also asked all four tools about this season's dates, hours and prices at 10 well-known haunted houses and theme-park Halloween events, plus Spirit Halloween's store hours. Then we checked 46 answers against each attraction's official website.
Eight of them had at least one fact that was wrong or out of date. That's about 1 in 6. Some of the misses:
- "Not open yet" (it was): Copilot said Netherworld, near Atlanta, had not released its 2026 operating dates or ticket prices. Netherworld's site listed the full 2026 season, open since September 18, with tickets from $25 to $41.
- An old price: ChatGPT said single-night tickets to Knott's Scary Farm in Buena Park, Calif., start at $65. Knott's now lists them from $75. Copilot gave the same old $65 figure.
- The wrong hours: Google's AI Overview said Cedar Point's HalloWeekends frights start "around 6:00 PM" on Thursdays, Fridays and Saturdays. Cedar Point's nighttime frights run 8 p.m. to midnight on Thursdays and Fridays. Copilot made a similar mistake.
Error rates by engine:
- Copilot: 4 of 12 answers had an error (33%)
- ChatGPT: 2 of 12 (17%)
- Google AI Overviews: 1 of 10 (10%)
- Perplexity: 1 of 12 (8%)
Most misses weren't wild. They were start times off by 30 minutes to two hours, or a launch-week price that had since gone up. But small errors are exactly the ones that cost you. Nobody double-checks an answer that sounds reasonable.
We were also strict about what counted as wrong. We only flagged a claim when the official page clearly contradicted it. Seven answers couldn't be checked either way, and we didn't count those as errors. So 1 in 6 is the floor, not the ceiling.
For a haunted house that earns much of its year in a few weekends, a wrong start time isn't a small thing. It's a family pulling up to a locked gate at 6 p.m. Or worse, a family that never comes because AI told them you weren't open yet. See our guide to event and attraction marketing for more.
Why AI gets your hours wrong
Again, our test shows that AI gets things wrong, not exactly why each mistake happened. But the usual suspects are easy to find on most seasonal sites:
- The facts live in a picture. Hours and prices posted as a flyer image, a social graphic or a PDF are much harder for AI tools to read than plain text.
- The facts live in a widget. Many ticketing and booking tools load prices with scripts after the page opens. Some AI crawlers don't see that content at all.
- Last year's page is still out there. If your 2025 hours page is still live and doesn't say which year it's for, AI may grab it instead of this year's.
- Other sites have old information. Local event roundups, directories and blog posts often list launch-week prices and never update them.
- Your listings don't match. If your website says 8 p.m. and your Google Business Profile says 7 p.m., AI has to pick one. It may not pick the right one.
What a good "season page" looks like
The fix is boring, but it's fast. Every seasonal business should have one clear page that answers the questions people ask AI. It should include, in plain text:
- The season and year in the headline (for example, "2026 Season Dates, Hours and Tickets")
- Opening and closing dates for the season
- Hours for each day you're open, including any days that differ
- Current ticket or service prices, and what changes them (weekends, peak nights, online vs. at the door)
- Your address and parking details
- Any closures, sell-outs or weather policies
- A "last updated" date
Keep it on one page that stays at the same web address every year, and update it each season instead of making a new one. That way the links and history you've built up keep working for you. Adding organization schema to your site can also help search engines and AI tools understand who you are.
How to run your own AI check in 20 minutes
You don't need special tools to see what AI says about you. Open ChatGPT, Perplexity, Copilot and Google and ask each one the same questions your customers would ask:
- "What are [your business]'s hours this season?"
- "How much are tickets to [your business] in 2026?"
- "Is [your business] open yet?"
- "Best [what you sell] near [your city]"
- "Where should I buy [your main product]?"
Write down every wrong fact, every missing mention and every competitor that shows up instead of you. Then look at which websites each answer cites. Those are the pages you need to fix or the sites you need to be on.
One check isn't enough, though. AI answers change from day to day and can differ by location. Run the same questions again every few weeks during your season.
This isn't just a Halloween problem
Swap "haunted house" for a ski rental shop, a tax preparer in April, a pumpkin patch, a holiday market, a summer camp or a pool supply store in May. Any business with a short busy season has the same risk. AI answers questions about your hours, prices and whether you're open, whether or not you've given it good information.
And the shopping results apply to any specialist. If you're the expert in one category, don't assume AI knows it. The big general stores are showing up in your aisle, one product question at a time. Our Halloween marketing glossary entry has more ideas for seasonal brands.
Before your busy season: a checklist
- Two months out: ask the AI tools about yourself. Ask ChatGPT, Perplexity, Copilot and Google for your dates, hours and prices. Write down what's wrong.
- Six weeks out: put the facts in plain text. Publish this season's dates, hours and prices on one crawlable page. Don't rely on images, PDFs or ticketing widgets alone.
- Label the year. Say which season the information is for ("2026 hours") so AI doesn't mix it up with last year. Update or redirect old pages.
- Fix your listings. Make sure your Google Business Profile and other local listings match your website, including closed or moved locations.
- Give each product its own page. If you sell decorations, candy or pet costumes too, make clear pages for them. Don't let one category define your whole brand.
- Earn the link, not just the mention. Show up where people talk about you, like Reddit, YouTube and local roundups. Ask the sites listing old prices to update them.
- Check again once the season starts. Run your AI check every few weeks. Answers change, and so do your prices.
About the data
Butter Labs ran 34 Halloween shopping prompts once in each of ChatGPT, Perplexity, Google AI Overviews and Copilot (US) on October 24, 2026, tracked with Otterly.AI. Google showed an AI Overview for 13 of the 34 questions, for 115 shopping answers in total. We tracked 28 retailers and candy makers. A brand counted once per answer if it was named, and "linked" means the answer cited the brand's own main website.
The attraction check covered 12 questions about 2026 dates, hours and prices, for 46 answers in total. Each claim was compared with the official website on October 24, 2026. We only called a claim wrong when the official page clearly contradicted it; 7 answers couldn't be verified and weren't counted as errors.
AI answers change from day to day and by location, so treat this as a snapshot, not a ranking. The attraction sample is small.
Read the full findings, charts and method in The Haunted AI Search Report.
Is AI ghosting your brand?
Butter's GEO plans start at $399/month and include monthly AI prompt testing, so you can see what ChatGPT, Perplexity, Copilot and Google say about you before your busy season, not after.
