"The fake references ChatGPT made up for you": why Consensus's Android app is getting two thousand students a day

Consensus is an AI academic search company that has raised nearly $50M. It only added an Android app in August 2026 and has gotten 144k installs in three months, mostly from students in the Philippines, Indonesia, and Nigeria. It's not a newly emerged product. An established brand has a new entry point. Below: where these installs come from, how it makes money, who it's competing with, and why we don't recommend a head-on fight, only copying a few of its moves.

A Malaysian lecturer's video

In Sep 8 2026, Malaysian university lecturer Dr Luqman Shah posted a video on TikTok with a title in Malay:

Risau dapat rujukan PALSU dalam ChatGPT? Gunalah kaedah ini! (Worried about fake references from ChatGPT? Use this method!)

Tags were #chatgpt #consensus #plugin. By Nov 6 it had 420k views. His account consistently covers one thing: how to write journal papers and get past Turnitin and AI detection. Several videos about "my AI detection rate is too high, what do I do" also sit around 400k views.

The same pitch is everywhere in Southeast Asia. An Indonesian account, @pande.digital, posted in Nov 1: "Using ChatGPT to find literature? Careful, the citations may be made up. Writing your thesis (skripsi)? Try these 3 AIs," with Consensus first. @fitrimarisachannel's June post was called "AI keeps hallucinating references? Try ChatGPT + Consensus."

One month ago, on Aug 5, Consensus launched its Android app on Google Play. On Oct 8, a Nigerian user named Adebayo left a review: "The info this app gives is mostly accurate and very useful for research, but I'm giving two stars because every time I open it, it kicks me out to the browser and I have to log in again. I want the app, not the browser." 20 days later, the official reply said it was a bug and had been fixed.

What it is

In one line: an AI search that only searches peer-reviewed papers. You ask a question, and it finds answers across 400M+ papers, writes a plain-language summary, and attaches a source paper to every sentence so you can click through and verify. For yes/no questions, it produces a "Consensus Meter" showing how many studies support, oppose, or reach no conclusion.

One line in the store description is the foundation of the whole business: "We partner with 7 of the top ten academic publishers": Wiley, AAAS, Sage, APA, ACS, giving access to full text behind paywalls you can't read elsewhere.

The app does what the web version does: search, read and ask follow-up questions on single papers, upload PDFs, ask by voice, export citations in APA / MLA / BibTeX, and import libraries from Zotero. You log in with your web account, and saved items and history sync over.

Where these installs come from

Look at the curve first. Google Play shows 144k cumulative installs (11-06), with 17.7k more in the last 7 days. By third-party daily estimates, it was 1,400–1,500 a day in mid-August, then stayed between 1,200–2,500 a day. Early September: 20k; early October: 60k; mid-October: 100k. There's no step, no single day of a sudden spike. It's not like a free-promotion chart push or a single big ad buy. It looks like a group of people who were already using it gradually switching to mobile.

It has only charted in the Education free charts of three countries: the Philippines (best rank #90), Indonesia (#145), and Nigeria (#193). It doesn't appear on charts in big markets like the US.

Installs come from three main sources:

There are other routes aimed at brands rather than direct Android installs: since 2023 it has been a top research GPT in the ChatGPT GPT Store; university libraries such as Yale and WashU list it as a recommended tool; podcast ad reads ("one month free trial"); and an abundance of "Elicit vs Consensus vs SciSpace" comparison articles.

One new change to watch: on TikTok in 2026, videos about "Claude + Consensus" have multiplied. A Spanish-language account, @carlitosmanosalvas, posted in July a video on connecting Consensus to Claude for research, with 80k views; an Indonesian account, @dr.mirzazaka, said in August, "Everyone says don't use Claude, the references hallucinate, but connect it to PubMed or Consensus and it's different." Its May 2026 launch week pushed an MCP interface and a new API. Its entry point is moving into other people's chat assistants; the mobile app is just one of the doors.

Where the money comes from

The app itself doesn't charge, and the store page doesn't list in-app purchases. The description lists two tiers, Pro (20 papers per synthesis) and Deep (searches 50 papers and produces a structured literature review), with subscriptions paid on the web. The official help page lists prices: Pro is $20/month or $144/year; Deep is $65/month or $540/year. There is also an institutional version for universities and teams.

Public revenue figures: in 2024 its Series A came with annualized revenue of $1.5M; in July 2025 ARR was $4.8M (Tim Draper said so on LinkedIn); in May 2026 it raised a $30M Series B. How much of that comes from mobile subscriptions isn't visible from outside.

Putting the subscription on the web and in the app for free saves the 15–30% store cut. For a company that already takes payment on the web, that's the smoothest move.

Who it's competing with

Interestingly, on Android it has almost no real competition. Elicit and SciSpace never shipped a proper official app, and students searching their names install a name-squatting "guide" app. That leaves Consensus an open lane on phones, but it's held by the brand, not by how good the mobile version is. The store reviews complain about slowness, panels that won't open, and constant bounces to the browser.

Can you follow it?

Hard to compete head-on. Most of what made it take off is out of reach for followers:

• Publisher licensing. Full-text paywalled content from 7 major publishers, the result of a company negotiating for years. A small team can't get those deals.
• 2.5M monthly web users. The Android installs are this audience switching devices, not new users drawn in by an app.
• It's already well known. Three funding rounds have been reported, revenue is public, reviews are everywhere, and when students search "AI for finding papers" it's on the first page.

Three things are worth copying:

1. Use a pitch built on something users truly fear. Students in Southeast Asia are less afraid of "can't find papers" than of "the references my submission cites were made up by ChatGPT." Nearly every viral video title uses that line.
2. Hire local lecturers, not influencers. A lecturer teaching "thesis writing tips" and bringing in the tool along the way gets trust from students at low cost. Academic accounts in Malaysia, Indonesia, Vietnam and Nigeria often pull tens of thousands to hundreds of thousands of views.
3. Web first, app second, and collect payment on the web. The app only catches existing users; it doesn't carry the payment.

If it were you

C

There are lessons here, but following it isn't worth much. Copying an "AI that only searches papers" to fight Consensus means going up against it with no publisher content, no existing users, and no name. Its free tier includes a set number of deep searches a month, which is enough for most students (inference).

Who it suits: people who already have voice in a student community in a specific country, such as a lecturer or PhD with an academic account in Indonesia, Vietnam or Nigeria, who want to build a local-language "thesis assistant." It's not a search engine but an end-to-end flow: finding real literature, formatting citations to the local university's style, and self-checks before plagiarism and AI-detection scans. Consensus only covers the first step, and its store page and blurb are English-only (whether the interface has local languages is unverified).

Who it doesn't suit: anyone aiming at the global English market, anyone hoping to build academic search more accurate than Consensus with tech alone (accuracy depends on the paper library and licensing, not the model), or anyone without a local student channel who only writes code.

How to build it: use public paper APIs for retrieval (Semantic Scholar, OpenAlex and Crossref are all free), or connect to Consensus's own open API, and don't build your own index. Focus on the workflow and localization: questions in local languages, citation formats for local universities, and templates for each thesis chapter. Take payment with local payment methods and cheap per-semester plans, not a copy of the $20 monthly price (inference: students in these markets have a much lower ceiling on what they'll pay).

How to get users: follow Consensus's approach in Southeast Asia. Have local lecturers film videos like "ChatGPT's references are fake, here's how to check," one pain point per video: fake citations, Turnitin similarity scores, AI-detection rates, and papers sent back by supervisors over formatting.

Timing: aim for thesis season. In Indonesia and the Philippines, end-of-semester thesis submission peaks roughly in 12–January and 5–July (inference, varies by school), so have a usable version and a few videos running by April of 2027 at the latest.

Signs you've got it wrong: (1) Consensus launches local-language interfaces such as Indonesian or Vietnamese, or signs whole-campus licenses in these countries and universities, and the localization gap is filled by them. (2) Google Scholar, Gemini or ChatGPT's free tier starts giving clickable, verified paper citations directly, and the "AI makes up papers" pain point disappears with it. (3) The reverse: if Consensus's Android installs keep charting in Indonesia and the Philippines and climb toward the top 50 in education, while no local-language equivalent exists yet, then this student demand is bigger than it looks, and it's worth moving sooner.

Last point: what Consensus's mobile version tells you isn't that "academic AI still has room." It's that a product already built on the web can add a port and new countries, and gain two thousand more users a day without spending much. If you have a web product that already works, ask yourself: where are my existing users, what phones do they use, and has anyone on TikTok already explained my selling point for me?

Data and sources

• Installs, ratings, charts: Google Play store page and install counts (as of 11-06). Daily installs are third-party estimates (AppstoreSpy). Chart is the education free chart for the Philippines, Indonesia and Nigeria (from 10-29).
• Store blurb and reviews, quoted: Google Play store page (11-06).
• Pricing: Consensus help center "Subscription plans" page; not verified by purchase.
• Funding and revenue: Consensus official blog (Series A, 2026-05 Series B announcements), arr.club (2025-07 ARR, sourced from Tim Draper's LinkedIn), Tracxn (about $48.9M cumulative).
• TikTok videos: titles, dates and view counts as seen in our search for 11-06; whether they are paid partnerships is unverified.
• Competitor scale: Google Play install counts, App Store US ratings count (around 11-06).
• iOS: Consensus's iOS app doesn't show up on the US App Store (11-06). Whether it's live in other storefronts is unverified.