What AI knows can change which businesses it recommends
AI can use what it knows about a customer to choose which businesses to recommend. Give it complete, accurate information about every location so it can tell when your business meets that customer’s needs.
1,761,754 answers · 384,076 locations · 6 answer systems
Q3 2026 observations beginning 1 July · Analysis prepared 31 August 2026
AI search · context · memory · complete profiles · customer needs
Each rate uses answers about businesses that list the feature as available.
A broad AI question gives the system little information about the customer. A need such as wheelchair access, vegan food or delivery can change which businesses fit. Give AI complete facts about each location so it can match the right business to the right person.
This study tracks the facts AI mentions in broad descriptions with no memory from earlier chats. Those answers show only one kind of question. They cannot tell a brand what every customer will need.
Customer needs change which facts AI may use
A broad hotel answer may lead with the pool. A guest traveling with a parent who has limited mobility may care more about the entrance, parking and route to the room. A low mention rate does not make those facts less useful to that guest.
Memory can carry a customer’s needs into a later question. The customer may not need to type the same detail again. New memory features from OpenAI, Anthropic and Google show that this is becoming common across AI assistants.
AI assistants can carry details into the next task
OpenAI · 4 June 2026. Its new ChatGPT memory system can use preferences and limits from earlier chats. OpenAI shows how this can shape later hotel and restaurant suggestions.
Anthropic · 25 August 2026. Claude now shares memory across chat and Cowork. A detail from a chat can help with a later task. Users can see and change what Claude remembers.
Google · 26 March 2026. Gemini added imports for memories and chat history from other AI apps. A person can bring past preferences and details into Gemini.
The latest question may include only part of what AI knows about a customer. A broad scan cannot show every fact that may matter later. Give AI complete location information so it has the right facts when a customer needs them.
A lunch answer uses a detail from an earlier chat
These chats took place a couple of days apart. The first asks about a walking stick for the customer’s grandfather. The second asks where to have lunch with him in New York. It does not mention the walking stick or ask about access.
The earlier chat
A couple of days later
“For lunch with your grandfather, I’d prioritize somewhere comfortable, easy to get into, not too loud, and with a broad menu.”
The lunch response says the restaurant should be easy to enter. The earlier chat explains why that detail matters. Clear facts about entrances, seating and access can help AI decide whether a restaurant fits before it chooses which businesses to name.
Brands cannot predict every question a customer will ask. They can give AI complete facts about each location. That includes details a broad answer may rarely mention.
The study asks AI to describe a named location. The prompt does not include memory from earlier chats. Each chart carries two numbers for every fact. Filled in is the share of locations offered that field which mark it Yes. Mentioned is the share of answers that mention the fact, counted only over locations where it is marked Yes.
The two numbers use different denominators and are not a before and after. Read them as what brands record against what AI says. A mention may say that a service is offered, unavailable or limited by a rule.
What a brand fills in does not predict what AI mentions
Every fact in this study has two numbers. The first is how often locations that are offered the field mark it Yes. The second is how often AI mentions it. If the two moved together the cloud would rise to the right. It does not.
Pool sits at the top left: 3.37% of the hotels offered the field mark it Yes, and pool is mentioned in 83.09% of the answers about hotels that do. Wheelchair accessible parking sits at the bottom right: 76.96% of locations mark it Yes and it is mentioned in 0.20% of answers. Filling a field in does not make AI say it, and a fact AI rarely says out loud is not a fact you can leave blank.
157 facts across five industries. Mention rates here are pooled across all industries and can differ by a few tenths from the per-industry rates in the charts below. One further fact is held out of this chart because its filled-in share cannot be computed against a clean denominator.
This is the reason the industry charts that follow carry both numbers. A mention rate on its own tells a brand what AI talks about in one broad answer. It does not tell a brand which fields to fill.
Almost every hotel fills in free Wi-Fi. It is the least mentioned fact here.
Free Wi-Fi is marked Yes by 97.98% of the hotels offered the field and mentioned in 55.44% of answers, the lowest rate on this chart. A fact almost every hotel has does not tell a guest much. The amenities AI leads with, pool and spa, are marked Yes by very few hotels.
Pool and fitness center lead the mentions and sit two tenths of a point apart, yet a pool is marked Yes by 3.37% of hotels and a fitness center by 72.85%. What AI says first is not what hotels record most.
Hospitality totals · 80,981 answers about 18,054 locations. Each fact has its own set of answers from locations where that fact is marked Yes.
See all measured facts for this industry
| Attribute | Filled in | Mention rate |
|---|---|---|
| Pool | 3.37% | 83.41% |
| Fitness center | 72.85% | 83.21% |
| Spa | 7.11% | 64.18% |
| Bar | 34.65% | 63.07% |
| Pet-friendly | 66.79% | 62.92% |
| Indoor pool | 26.82% | 59.50% |
| Outdoor pool | 28.92% | 57.47% |
| Airport shuttle | 13.71% | 56.63% |
| Free Wi-Fi | 97.98% | 55.44% |
| Business center | 50.45% | 48.15% |
| Hot tub | 31.10% | 44.79% |
| Free breakfast | 59.55% | 36.33% |
| Free parking | 75.60% | 35.43% |
Fewer than half of restaurants fill in breakfast. AI mentions it most.
Breakfast is marked Yes by 45.27% of the restaurants offered the field and mentioned in 68.37% of answers. Delivery runs the same way, 22.59% against 54.59%. These are fields AI volunteers and most restaurants have not recorded.
Casual atmosphere and dessert run the other way. Both are marked Yes by most restaurants and mentioned least on this chart.
Food and drink totals · 295,477 answers about 62,554 locations. Each fact has its own set of answers from locations where that fact is marked Yes.
See all measured facts for this industry
| Attribute | Filled in | Mention rate |
|---|---|---|
| Breakfast | 45.27% | 68.37% |
| Delivery | 22.59% | 54.59% |
| Upscale atmosphere | 1.24% | 28.06% |
| Bar games | 0.40% | 27.56% |
| Shows sports | 3.30% | 26.52% |
| Live music | 1.67% | 25.25% |
| Beer | 10.73% | 23.16% |
| Catering | 28.05% | 21.98% |
| Karaoke | 0.30% | 20.27% |
| Cocktails | 6.79% | 20.04% |
| Vegan options | 8.49% | 19.65% |
| Casual atmosphere | 78.16% | 15.45% |
| Dessert | 65.46% | 15.17% |
Almost no provider fills in walk-ins. AI mentions them in a third of answers.
Accepts walk-ins is marked Yes by 0.66% of the providers offered the field and mentioned in 35.95% of answers. That is not a third of providers taking walk-ins. It is AI treating the question as worth answering, whether or not there is a fact to answer it from.
Online care shows the same shape at a smaller scale. Parking runs the other way: a free parking lot is marked Yes by 20.48% of providers and mentioned in 2.66% of answers.
Healthcare totals · 333,196 answers about 72,734 locations. Each fact has its own set of answers from locations where that fact is marked Yes.
See all measured facts for this industry
| Attribute | Filled in | Mention rate |
|---|---|---|
| Accepts walk-ins | 0.66% | 35.95% |
| Online care | 6.73% | 25.17% |
| Emergency services | 14.80% | 11.64% |
| Payment plans | 27.75% | 9.62% |
| Delivery | 22.59% | 4.01% |
| In-store pickup | 73.23% | 3.09% |
| Free parking lot | 20.48% | 2.66% |
| On-site parking | 7.42% | 2.21% |
| Parking garage | 0.88% | 1.73% |
Delivery is mentioned in a third of retail answers. One retailer in five fills it in.
Delivery is marked Yes by 22.59% of the retailers offered the field and mentioned in 34.41% of answers. Bakery and repair services show the same shape. In-store pickup runs the other way: most retailers fill it in and AI rarely says it.
Curbside pickup is marked Yes by more than half of retailers and mentioned in a fifth of answers. Installation service runs the other way.
Retail totals · 163,680 answers about 37,336 locations. Each fact has its own set of answers from locations where that fact is marked Yes.
See all measured facts for this industry
| Attribute | Filled in | Mention rate |
|---|---|---|
| Bakery | 6.35% | 37.25% |
| Repair services | 9.97% | 36.53% |
| Delivery | 22.59% | 34.41% |
| Drive-through | 43.69% | 28.27% |
| Curbside pickup | 53.71% | 21.36% |
| Installation service | 6.80% | 18.29% |
| Small business | 0.66% | 18.07% |
| Buys used goods | 0.47% | 12.92% |
| In-store pickup | 73.23% | 11.05% |
| Restroom | 38.31% | 9.47% |
Half of financial locations fill in card payments. AI mentions them in one answer in twenty.
Credit cards are marked Yes by 53.48% of locations and mentioned in 5.27% of answers. Debit cards, 56.36% against 3.94%. Most places take cards, so saying so helps very few customers choose, and that is the pattern across this whole chart.
Wheelchair accessible entrance is the widest gap on this chart. It is marked Yes by 75.73% of locations and appears in 0.69% of answers.
Financial services totals · 183,664 answers about 43,412 locations. Each fact has its own set of answers from locations where that fact is marked Yes.
See all measured facts for this industry
| Attribute | Filled in | Mention rate |
|---|---|---|
| Credit cards | 53.48% | 5.27% |
| Debit cards | 56.36% | 3.94% |
| Online appointments | 2.79% | 2.27% |
| Women-owned | 1.25% | 1.58% |
| Appointment required | 7.33% | 1.18% |
| Wheelchair accessible entrance | 75.73% | 0.69% |
| On-site parking | 7.42% | 0.62% |
| Cash advance | 51.47% | 0.62% |
| Free street parking | 14.92% | 0.45% |
| Free parking lot | 20.48% | 0.36% |
| LGBTQ+ friendly | 10.08% | 0.23% |
| Wheelchair accessible seating | 13.35% | 0.20% |
Why broad AI answers cannot guide your whole content plan
A broad prompt does not include every customer need. It may miss the customer who needs an accessible entrance, the guest traveling with a dog or the shopper who needs delivery. Do not use one broad answer to decide which facts your brand can leave blank.
AI assistants can use details from current and past conversations. Brands need to be ready for needs that the latest question does not spell out. Publish every fact that applies, keep it accurate and explain any limits that may affect a visit.
Earlier Yext Research on specific questions and specific pages found that AI cites different pages when a question names a service. The next study should test what changes when AI also knows the customer’s needs.
Give AI every fact that applies to each location
Fill every field that applies
Fill every field that applies. Add an accurate Yes or No, plus the service, access and payment details customers may need.
Keep each detail current
Update services, hours, fees and access rules when they change. Make sure profiles and location pages give customers the same answer.
Test real customer needs
Test broad questions, specific needs and follow-up chats. Look for missing facts that may decide whether the location works for that customer.
Give AI complete facts about every location. What AI knows about the customer can shape how it describes your business and whether it recommends you.
Test how memory changes later answers
A new study should ask about the same locations for customers with different needs. It should then ask follow-up questions that do and do not use earlier chat history. The study can track which facts AI mentions, which businesses it recommends and whether the answer helps the customer plan a visit.
Questions readers ask
How can context and memory change an AI answer?+
Should we fill fields that AI rarely mentions?+
What does this study measure?+
Are broad AI scans still useful?+
What should we do when a location does not offer a service?+
How we measured the results
- Study design
- We studied prompts that name a location and ask AI to describe it. The prompt does not ask about a specific fact. The study includes 1,761,754 answers about 384,076 locations from 6 AI systems. The observations begin on 1 July 2026. We prepared the analysis on 31 August 2026.
- Answer systems
- The study covers Perplexity, Google AI Mode, Gemini, Claude, Google AI Overview and SearchGPT. Each answer is tied to a location, AI system, search term and type of prompt. The charts use only the broad prompt that names a location.
- Attributes
- The study starts with 178 fields that can be marked Yes or No. Of these, 47 apply across all five industries and 131 apply only to certain industries. The final analysis includes 158 facts. Each fact had at least 300 locations marked Yes. Opening hours are not part of this group.
- Mention measurement
- For each fact, we looked only at answers about locations where that fact is marked Yes. We then counted answers that used words linked to that fact. We left out questions that already named the fact. The word match may miss other ways of saying the same thing. A mention may say that the service is offered, unavailable or limited.
- Industry groups
- The search term places each answer in an industry. The report covers food and drink, healthcare, retail, hospitality and financial services. The total answers and locations show the size of each industry group. Each fact rate uses a smaller set of answers about locations where that fact is marked Yes. The full tables list every fact shown for each industry.
- Filled-in measurement
- The filled-in share uses the locations that are offered a field at all, not every location in the study. Hospitality has 26,344 such locations, and the answer study has 18,054 hotel locations. The two numbers on a chart therefore rest on different denominators and should not be subtracted from one another.
- Category assignment
- For the filled-in share, the fields in a location record identify which locations are offered a fact. For the AI answers, the search term sets the industry. We keep these two groups separate. We leave out clinical fields when the field itself is what places a location in the group.
- Interpretation
- The mention rates show which facts appear in broad AI answers. The filled-in shares show what brands record. The report does not measure what happens after a brand fills a field, and the weak fit between the two numbers is not evidence that filling a field has no effect.
- Context and memory
- The study prompts do not use memory from earlier chats. The walking-stick and lunch example shows how an earlier detail can appear in a later answer. The charts cover broad questions. Our advice to complete every field that applies prepares brands for more specific customer needs.