Beket AI beket.ai
Cover: Does AI recommend different apps to a founder, a CFO and a parent? Study date November 17, 2026. Cards show 64 of 72 answers recorded, Monarch Money first in 9 of 22 parent answers, and the same first app in 2 of 19 CFO vs parent cases.

Does AI Recommend Different Apps to a Founder, a CFO and a Parent? 2026 Study

By Beket AI team

Study date: November 17, 2026

We asked ChatGPT, Perplexity and Google AI Overview the same 8 budgeting-app questions three times each, once as "I'm a startup founder.", once as "I'm a CFO." and once as "I'm a parent." The person asking changed the answer. Monarch Money was the first app named in 9 of the 22 parent answers. For the CFO, Ramp was first most often, in 4 of 21, and some CFO answers moved to finance software for companies. When we compared the CFO and parent answers to the same question on the same engine, the first app matched in only 2 of 19 cases.

In Are ChatGPT answers the same for everyone? we covered what OpenAI and Google document about personalization, and how a role stated in the question can shift a recommendation. This is our own small test of that idea. We picked budgeting and spending-tracker apps because a founder, a CFO and a parent all really ask about them.

These numbers are what the AI engines returned. They are not Beket AI's opinion of any app. AI answers change from run to run, so read this as one snapshot, not a measured effect.

Study detail Value
Study date November 17, 2026 (run about 5:20 PM to 5:27 PM PT)
Topic Budgeting and spending-tracker apps
Engines ChatGPT, Perplexity, Google AI Overview (one audit)
Questions 8, none naming an app
Persona lines 3, each typed at the start of the question
Language and location English, no location set
Answers recorded 64 of 72 possible
Missing answers 8, all from Google AI Overview

The 8 questions and the 3 persona lines

Each question was run three times, with one of these lines in front of it:

  • "I'm a startup founder."
  • "I'm a CFO."
  • "I'm a parent."

For example, one prompt read "I'm a CFO. What's the best free budgeting app?" The 8 questions, with none naming a brand:

  1. What's the best app to track spending and stick to a budget?
  2. Which budgeting app is actually worth paying for in 2026?
  3. What's the best free budgeting app?
  4. What app should I use to track receipts and expenses?
  5. What's the best app for managing shared expenses with other people?
  6. What's the best tool to see all my bank accounts and cards in one place?
  7. Which app helps me plan for big upcoming expenses and savings goals?
  8. What's the most secure money app to connect my bank accounts to?

That makes 24 prompts and 72 answer slots across the three engines. We did not run a version without a persona line, so we cannot say how any persona differs from a plain question. We compare the three personas with each other.

How many answers did we get?

ChatGPT and Perplexity answered all 24 prompts each. Google AI Overview answered 16 of 24.

Engine Founder CFO Parent Total
ChatGPT 8 of 8 8 of 8 8 of 8 24 of 24
Perplexity 8 of 8 8 of 8 8 of 8 24 of 24
Google AI Overview 5 of 8 5 of 8 6 of 8 16 of 24
All engines 21 of 24 21 of 24 22 of 24 64 of 72

Missing Google AI Overview answers

Google AI Overview returned no answer for 8 prompts. These are missing, not zero. The audit run reported all 72 prompts as landed, but these 8 answers have the status "not generated," and the audit's own count of answers is 64. Here is what is missing:

  • Founder: question 1 (spending and budget), question 5 (shared expenses), question 7 (upcoming expenses and savings goals)
  • CFO: question 6 (all accounts in one place), question 7 (upcoming expenses and savings goals), question 8 (most secure money app)
  • Parent: question 2 (worth paying for), question 6 (all accounts in one place), question 7 (upcoming expenses and savings goals)

One more AI Overview answer exists but named no app: the parent's shared-expenses answer (question 5). It counts as an answer with no app named, not as a missing one.

Which app came first for each persona?

We counted the first app named in each answer. Across all three engines:

Persona Answers Most often first Next most often first
Founder 21 YNAB (5) Ramp (2)
CFO 21 Ramp (4) YNAB (3), Monarch Money (2), Quicken (2)
Parent 22 Monarch Money (9) Goodbudget (2), YNAB (2)

Source: Beket AI market audit, November 17, 2026. Every other app was first once at most.

The parent answers were the most consistent. Monarch Money led 9 of 22, and it led 3 times on each of the three engines. The founder and CFO answers were spread across more different first apps.

Who got which apps?

Next we counted how many answers named each app at all, across all engines. Each persona has 21 or 22 answers, so the counts are comparable.

App Founder (of 21) CFO (of 21) Parent (of 22)
Monarch Money 9 8 14
YNAB 7 9 10
Quicken Simplifi 4 7 7
QuickBooks 9 5 0
Ramp 6 6 0
Xero 8 3 0
Brex 6 2 0
Google Sheets 3 5 0
Excel 2 4 0
Rocket Money 2 2 6
Goodbudget 2 1 6
EveryDollar 2 2 5

Source: Beket AI market audit, November 17, 2026. Counts are answers that named the app, not mentions.

Grouped bar chart of how many answers named each app for a founder, a CFO and a parent. Monarch Money: 9, 8 and 14. YNAB: 7, 9 and 10. Quicken Simplifi: 4, 7 and 7. QuickBooks: 9, 5 and 0. Ramp: 6, 6 and 0. Xero: 8, 3 and 0. Rocket Money: 2, 2 and 6. Goodbudget: 2, 1 and 6. EveryDollar: 2, 2 and 5.

Three patterns stand out:

  1. A shared core. Monarch Money and YNAB were named for every persona, in 8 to 14 and 7 to 10 answers. Quicken Simplifi was named for all three too, in 4, 7 and 7.
  2. Business tools stopped at the parent. Ramp, QuickBooks, Xero and Brex were named in answers for the founder and the CFO but in none of the 22 parent answers.
  3. Family and low-cost apps rose for the parent. Rocket Money was named in 6 parent answers against 2 each for the others. Goodbudget was named in 6 against 2 and 1. EveryDollar was named in 5 against 2 and 2.

Some apps were close to one persona only. Wave (4 answers) and Mercury (3) appeared only for the founder. NetSuite (4), Tiller (3), SAP Concur (3) and Sage Intacct (3) leaned to the CFO, and Google Sheets and Excel were named in 5 and 4 CFO answers against 3 and 2 for the founder and none for the parent.

How different were the answers, persona to persona?

For every question and engine where two personas both had an answer, we asked whether the first app was the same, and whether the top three apps shared any name. "Top three" means the first three different apps in the order they appear in the answer text.

Persona pair Comparable cases Same first app Top three share an app Top three share no app
Founder vs CFO 19 8 15 4
Founder vs parent 19 5 11 8
CFO vs parent 19 2 8 11

Source: Beket AI market audit, November 17, 2026. A case is one question on one engine where both personas have an answer.

Two bar panels. Left: how often the first app was the same for a pair of personas. Founder vs CFO 8 of 19, founder vs parent 5 of 19, CFO vs parent 2 of 19. Right: how often the first app was the same across persona pairs on each engine. ChatGPT 11 of 24, Perplexity 2 of 24, Google AI Overview 2 of 9.

The parent was the most different from the other two. The CFO and the parent got the same first app in 2 of 19 cases, and their top three apps shared no name in 11 of 19. The founder and the CFO were the closest pair, with the same first app in 8 of 19.

Did the CFO prefix change the kind of tool?

Sometimes it did. A few CFO answers moved from consumer budgeting apps to finance software for companies, while the founder and the parent got consumer budgeting apps for the same question.

  • Worth paying for (question 2), ChatGPT: the CFO answer named Aleph first, then Excel and Google Sheets, and later Cube and Datarails. The founder and parent answers led with YNAB.
  • Upcoming expenses (question 7), ChatGPT: the CFO answer began with Monarch Money, YNAB and Quicken Simplifi, then added Planful and Pigment.
  • All accounts in one place (question 6): the CFO answers named Trovata on ChatGPT and Kyriba on Perplexity. The parent answers named Monarch Money first on both.
  • Receipts and expenses (question 4), ChatGPT: the CFO answer named NetSuite, SAP Concur and Microsoft Dynamics among its apps.

The parent prefix also moved some answers toward family and money-movement tools:

  • Shared expenses (question 5): Perplexity named Onward, Venmo, PayPal, Zelle and Cash App. ChatGPT named Splitwise, then OurFamilyWizard and Settle Up.
  • Receipts and expenses (question 4): Perplexity named SupportPay first.
  • Security (question 8): Perplexity named Apple and Apple Cash Family first.

Did the three engines behave the same way?

No. The persona line mattered very differently depending on the engine.

Engine Answers recorded Same first app across persona pairs Top three share no app
ChatGPT 24 of 24 11 of 24 5 of 24
Perplexity 24 of 24 2 of 24 14 of 24
Google AI Overview 16 of 24 2 of 9 4 of 9

Source: Beket AI market audit, November 17, 2026. Each cell pools the three persona pairs. Google AI Overview has few comparable cases because of the missing answers.

  • ChatGPT was the steadiest. For 2 of the 8 questions, shared expenses (question 5) and the most secure app (question 8), it gave the same first app to all three personas. The first app matched in 11 of 24 pair cases. Its most common first apps were Ramp and YNAB for the founder (2 answers each), Ramp and Monarch Money for the CFO (2 each) and Monarch Money for the parent (3).
  • Perplexity changed the most. It never gave all three personas the same first app on any of the 8 questions. The first app matched in only 2 of 24 pair cases, and the top three shared no app in 14 of 24. Its most common first apps were YNAB for the founder (2), Quicken for the CFO (2) and Monarch Money for the parent (3).
  • Google AI Overview answered 5, 5 and 6 questions for the three personas. Monarch Money was first in 3 of the 6 parent answers. For the founder and CFO, no app was first more than once. Only 2 of its 9 pair cases matched on the first app. With so few answers, this is the least reliable column.

If you checked only one engine, you could reach a different conclusion about persona than someone who checked another.

First app, by question, persona and engine

This is the first app named in each answer. "Missing" means the engine returned no answer. These are not zeros.

ChatGPT

Question Founder CFO Parent
1. Track spending, stick to a budget YNAB YNAB Monarch Money
2. Worth paying for in 2026 YNAB Aleph YNAB
3. Best free app Empower Empower Goodbudget
4. Receipts and expenses Ramp Ramp Expensify
5. Shared expenses Splitwise Splitwise Splitwise
6. All accounts in one place Ramp Ramp Monarch Money
7. Upcoming expenses, savings goals Quicken Simplifi Monarch Money YNAB
8. Most secure app Monarch Money Monarch Money Monarch Money

Perplexity

Question Founder CFO Parent
1. Track spending, stick to a budget YNAB YNAB Monarch Money
2. Worth paying for in 2026 YNAB Quicken Monarch Money
3. Best free app Wave GnuCash Goodbudget
4. Receipts and expenses Expensify Ramp SupportPay
5. Shared expenses Tricount Expensify Onward
6. All accounts in one place Puzzle Kyriba Monarch Money
7. Upcoming expenses, savings goals Quicken Quicken Savemo
8. Most secure app Stripe Chase Apple

Google AI Overview

Question Founder CFO Parent
1. Track spending, stick to a budget Missing YNAB Monarch Money
2. Worth paying for in 2026 YNAB Tiller Missing
3. Best free app EveryDollar Google Sheets EveryDollar
4. Receipts and expenses QuickBooks Ramp Monarch Money
5. Shared expenses Missing Tricount No app named
6. All accounts in one place Mercury Missing Missing
7. Upcoming expenses, savings goals Missing Missing Monarch Money
8. Most secure app Plaid Missing Plaid

Source: Beket AI market audit, November 17, 2026. The first app is the first one the audit found in the answer text.

What this means for a brand

These are things the data supports, not a promise of results:

  • Test the roles your buyers have. In this run, an app named for a founder was sometimes absent for a parent. Ramp, QuickBooks, Xero and Brex were named for the founder and the CFO and never for the parent.
  • Check each engine on its own. Perplexity moved with the persona far more than ChatGPT did. A single-engine check can understate or overstate the effect.
  • Expect a different category, not only a different order. Some CFO answers switched to finance planning, treasury or ERP software. A brand that sells to consumers may be missing from a business-role answer, whatever its rank elsewhere.
  • Treat one run as a direction, and repeat it. Our persona prompts guide explains how to write role-based test questions, and why checks run through an API don't always match what users see is worth reading before you compare numbers across tools.

Limitations

  • Single run. All the answers come from one run on one day, with no repeats. Differences of one or two answers are within normal run-to-run variation. As we showed in our explainer on ChatGPT personalization, lists vary even when the prompt is identical. Treat these results as direction, not as measured effect sizes.
  • A text prefix, not a real user. The persona is a sentence at the start of the question. It is not a real account, ChatGPT Memory or a Google personalization profile. This tests how a stated role changes the answer, not what a real person with their own history would see.
  • Eight questions. We used 8 questions, not a larger set, so one question can swing a result.
  • A thin Google AI Overview base. 8 of 24 AI Overview answers are missing. Its comparisons rest on 5 or 6 answers per persona and only 9 comparable cases. Its pair figures should not be weighed like the other two engines'.
  • No control. There was no run without a persona line, so we cannot tell which persona is closest to a plain question.
  • Names come from the audit. Names are the apps the audit's entity extraction found in each answer. We merged obvious aliases (for example Monarch and Monarch Money, YNAB and You Need A Budget, Simplifi and Quicken Simplifi, and QuickBooks Online and QuickBooks), and we dropped labels that are not products, such as publishers, people, forums and operating systems. Quicken and Quicken Simplifi stay separate. Spreadsheets, banks and payment tools stay in because the engines named them. A different reading of the answers could give slightly different counts.

Methodology

On November 17, 2026, between about 5:20 PM and 5:27 PM PT, we ran one Beket AI market audit on the topic "budgeting and spending-tracker apps." It had 24 prompts: the 8 questions above, each with one of the 3 persona lines at the start. It ran on ChatGPT through the audit's ChatGPT app connector, on Perplexity and on Google AI Overview, in English with no location set. We turned off prompt generation, so only these 24 prompts ran.

The audit reported 72 of 72 answers as landed. 8 Google AI Overview answers had the status "not generated," so we counted 64 answers and listed the 8 as missing. For each recorded answer we took the apps the audit identified, removed repeats within the answer, and kept their order of first appearance. "First app" is the first of that list. "Top three" is the first three. Counts of answers naming an app are counts of distinct answers. They do not match the audit's market leaderboard, which counts every mention.

Nothing here is advice on which app to use. If you want to see how AI describes your own business, you can run a free audit at beket.ai.

Related posts