RETROSPECTIVE RECORD · PREPARED 16 SEPTEMBER 2026The archive · 100 retrospective records ↗
Lovebot Journaljournal
← The archive

Analyst estimates of companion app spending are not disclosures

Two analyst firms' own posts show how download and spend estimates are built, and why they are not company disclosures.

First source published
March 31, 2023
Site publication
September 18, 2026
Visual for this record: Analyst estimates of companion app spending are not disclosures
Visual published by macitynet.it, shown for identification of the record. Credit: macitynet.it · source page ↗ Rights: owner-review-pending. Source

What happened

Two mobile-analytics firms have published, separately, their own estimates of AI chatbot and companion app economics. Appfigures' post, dated 31 March 2023, estimated that three chatbot apps ranked in the App Store's top 50 — Kiyo, Genie and AI Chatbot — had together earned about $3 million in net revenue since the start of that year, averaging '$20K to $30K in revenue every day'. Sensor Tower's later report, covering January through August 2024, put Character.AI's downloads at 'nearly 19 million' over that period and its August 2024 monthly active users at 22 million.

What the documents show

Neither report names the same apps: Appfigures' 2023 post tracks generic chatbot apps that are not household names, while Sensor Tower's 2024 report tracks Character.AI specifically, describing it as the companion segment's 'frontrunner' and noting Talkie AI's downloads passed it inside the United States. Both explicitly label their figures as estimates. Appfigures states its numbers are 'net' revenue, meaning after Apple's and Google's store fees, drawn from its own App Intelligence platform. Sensor Tower states its revenue figures are 'estimated in-app purchase (IAP) estimates from the App Store and Google Play', excluding advertising revenue and third-party Android stores in mainland China.

The mechanism

An analyst estimate is built by modelling store-rank and download signals against a proprietary model, not by reading a company's ledger. That is why two firms, looking at the same broad category eighteen months apart, name different leading apps and never converge on one figure a reader could treat as ground truth. The mechanism worth naming here is the gap between an estimate and a disclosure: a company's own quarterly filing or investor letter states an audited number; an analyst post states a modelled one, and the model's assumptions — which stores, which countries, which revenue types — determine the result as much as the underlying app performance does.

What it leaves open

Neither post is a company disclosure, and no named app in either post has published a matching revenue figure of its own for comparison, at least not in the documents opened for this record. Sensor Tower's demographic breakdown — that users 18 to 35 account for over 70% of the segment, and that 66% of Character.AI's users are 18 to 24 — is likewise the company's own modelled estimate, not a survey of confirmed ages. Readers should treat every figure in this record as an estimate, clearly labelled as such by its publisher.

Both documents are useful precisely because they say what they are: modelled estimates with a stated method and period, not audited numbers. Treating them as anything more precise than that overstates what either publisher itself claims.

Sources & reading trail

Appfigures' own estimate of AI chatbot revenue in early 2023, with its stated method and scope.

Source published: 31 March 2023 · Retrieved: 16 September 2026

Sensor Tower's own estimate of Character.AI and companion-segment downloads, MAU and demographics for January-August 2024.

Source published: 1 September 2024 · Retrieved: 16 September 2026

Company documents, filings, studies and official records establish the record; the reading and the questions are Lovebot Journal editorial analysis. This retrospective draft does not imply the site published on the event date.

Continue reading

Sources & reading trail

The documents above establish the record. The reading and the questions are this publication’s editorial analysis, written after the fact.

Published September 18, 2026, not on the date of the event described.