Online therapy promised to fix mental healthcare's oldest problem — access. Then the most confidential conversation in medicine met the growth playbook.
Easier version: read this lesson at A2/B1 level →
Ten items. Work through the four activities — say your answer out loud before you reveal it.
Three terms: a controlled substance · a settlement · intake. Say the right one, then reveal.
Read each part. After it, there's a quick check (answer from memory first), then one question to discuss. The people in this story's user numbers were often at the hardest moment of their lives — the lesson keeps that in view throughout.
Begin with the part the later scandals made it easy to forget: the problem these companies set out to solve was, and remains, enormous. In most countries, mental healthcare is the queue that never shortens — waiting lists of months for a first appointment, rural regions with no therapist within a hundred kilometres, costs beyond most household budgets, and a stigma that stops millions from ever joining the queue at all. Estimates run to hundreds of millions of people worldwide with untreated depression or anxiety. Whatever else this case contains, "make therapy as easy to reach as a food delivery" was not a cynical pitch. For many users it worked: a counsellor within days instead of months, from a sofa instead of a waiting room, at a fraction of clinic prices.
The model that scaled was the marketplace: platforms such as BetterHelp and Talkspace signed up thousands of licensed therapists as flexible gig workers, matched them to subscribers by algorithm, and wrapped the whole encounter — video, voice, and above all unlimited messaging — in an app. Venture capital loved the shape of it: a vast underserved market, subscription revenue, marketing through the new intimacy channels — podcast hosts reading therapy ads between true-crime episodes, influencers describing their sessions. Then March 2020 detonated demand. Lockdowns created a mental-health emergency and simultaneously closed the clinics; regulators suspended old rules, including the requirement to see a prescriber in person; in some health systems, remote mental-health visits multiplied by an order of magnitude in weeks. Money followed: Talkspace reached the stock market in 2021 through a merger vehicle at a valuation around 1.4 billion dollars; newer entrant Cerebral raised hundreds of millions and passed a 4.8-billion-dollar valuation within two years of founding, with celebrity ambassadors and ads across every feed.
Inside the sector, two professional cultures were now sharing one product. Therapy's culture: fifty minutes, strict confidentiality, progress measured in months, success possibly meaning the patient leaves. The growth culture: acquisition funnels, conversion rates, engagement dashboards, retention. For a while, the tension stayed invisible — growth was access, access was mission, and the mission was going magnificently.
1. What real problem did the industry address, and what did it genuinely deliver?
2. What did the pandemic change — in demand, in rules, in money?
1. Untreated mental illness at global scale — waiting lists, therapy deserts, cost, stigma. The platforms genuinely lowered barriers: care in days not months, cheaper than clinics, reachable from home.
2. Demand exploded while clinics closed; regulators suspended in-person requirements (including for prescribing); remote mental-health visits multiplied; and capital flooded in — a ~$1.4bn SPAC listing for one platform, a ~$4.8bn valuation within two years for another.
The collisions arrived in quick succession, each along a different fault line. The first was data. In March 2023 the US Federal Trade Commission announced action against BetterHelp: the platform, it found, had shared users' health information — email addresses, IP addresses, and answers from the intake questionnaire, including the fact of seeking therapy — with advertising platforms including Facebook and Snapchat, to target ads and find similar prospective users, despite repeated promises of confidentiality. The settlement was modest in money — 7.8 million dollars, returned to users in the regulator's first-ever refund for mishandled health data — and enormous in precedent: the intake form, the most sensitive document in the funnel, had been treated as marketing data. The company, which admitted no wrongdoing and called the practices industry-standard, was banned from doing it again — and "industry-standard" was, investigations suggested, uncomfortably accurate: researchers found trackers throughout the mental-health app sector, and a separate scandal had already shown a crisis text service sharing anonymised conversations from suicidal moments with a commercial spin-off it owned.
The second reckoning was prescribing. Cerebral had grown at startup speed on a simple flow: online questionnaire, video assessment, medication by mail — including, under the suspended rules, controlled stimulants for ADHD. Reporting through 2022 described clinicians pressured toward thirty-minute evaluations and high prescribing rates, and nurses alleging targets that medicine should never have; pharmacies including CVS and Walmart began refusing the platform's stimulant prescriptions; federal prosecutors opened an investigation into possible violations of controlled-substances law; the founder-CEO was removed, and the company largely stopped prescribing controlled medications. No court ever ruled the model criminal — the investigation's existence, and the pharmacies' verdict, did the reputational work regardless. The VW mechanism, transplanted: no memo orders harm; the targets, the queue and the funding round order it obliquely.
The third reckoning was financial gravity. Talkspace, public at 1.4 billion, missed its numbers within months; both co-founders departed in late 2021; the stock lost on the order of ninety per cent, becoming the sector's WeWork-shaped exhibit — priced as a technology story, repriced as a healthcare-services company with therapist costs that scale with every patient. Across the sector, the pattern repeated in miniature: the pandemic cohort churned, marketing costs ate margins, and the dashboards that had justified the valuations began reporting the wrong direction.
1. What did the FTC find, and why did a $7.8m settlement matter far beyond its size?
2. What happened at Cerebral, and what mechanism does the lesson compare it to?
1. Intake answers and identifiers had been shared with ad platforms for targeting, despite confidentiality promises. The settlement set a precedent — the first FTC money returned for health-data sharing — and exposed the practice as sector-wide, not one bad actor.
2. Rapid online prescribing of controlled ADHD medication under suspended rules; reports of pressured 30-minute diagnoses; major pharmacies refused its prescriptions; a federal investigation followed; the CEO was removed and controlled prescribing largely stopped. The mechanism: VW's — no explicit order, targets doing the ordering.
What stabilised is more interesting than a simple collapse. Teletherapy itself is not in question: the evidence, across many trials, is that video-delivered therapy works about as well as the in-person kind for a wide range of conditions, and remote care has settled in as a permanent layer of mental healthcare — including inside sceptical public systems. The corrected model is quieter than the boom's: platforms integrating with insurers and employers rather than selling subscriptions off podcasts, regulators re-tightening prescribing, privacy treated as a survival requirement rather than a legal nuisance. The access gains were real and largely kept; the growth-at-intimacy's-expense playbook is what died — or at least went quiet.
Except that the next wave arrived before the last one finished breaking. AI chatbots now offer the conversation itself — always available, effectively free, infinitely patient, and answerable to no licensing board. The early record contains both signals: structured, boundaried programmes showing measurable benefit for mild conditions and gigantic unmet demand — and darker entries, from a peer-support platform revealed to have tested AI-written responses on users unaware they were talking partly to a machine, to lawsuits alleging that companion chatbots harmed vulnerable teenagers. The industry's accountability question, familiar from the translation case, arrives here with the stakes at maximum: an unaccountable system is cheapest precisely where the cost of error is highest.
Which leaves the case's living question, and it is not "was online therapy good or bad?" — the record answers both, measurably. The question is structural, and every healthcare system, employer and investor is answering it right now, mostly by default: the most confidential conversation in medicine is now mediated by companies whose economics reward scale, engagement and data — so who gets to set the boundaries the therapist's office once set by its walls: the market that funds access, the professions that carry the duty, or the regulators who arrive, as this case shows, roughly three years after the funnel? Somewhere tonight a person types their worst thing into a text box. The whole case is about everything that happens next to those words.
1. What survived the reckonings, and on what evidence?
2. What does the AI wave change about the case's central question?
1. Teletherapy itself: trials show video therapy roughly matches in-person outcomes for many conditions, and remote care is now a permanent, quieter layer — insurer-integrated, better regulated, with the access gains largely kept.
2. Chatbots deliver the conversation with no licensed human at all — cheapest exactly where error costs most. The accountability fortress from the translation case is being tested at maximum stakes, with early evidence pointing both ways.
This case demands precision about which data, which users, which promises — and judgement delivered in the same breath as fact. Relative clauses are the instrument for both, and C1 is where you gain their full range.
Insert the clause as defining or non-defining as the bracket instructs, and say aloud what each version claims.
1. The apps (that/contain trackers — defining) were flagged by researchers.
2. The intake questionnaire (which/every user completes — non-defining) became the centre of the case.
3. Rewrite with whose: The company's economics rewarded engagement. It faced a conflict. → The company, , faced a conflict.
1. that contained trackers (no commas — only those apps)
2. , which every user completes, (commas — there is one questionnaire; the clause adds information)
3. whose economics rewarded engagement
Rewrite as indicated. Write the full sentence.
1. Quantifier pattern: The users were mostly seeking help for the first time. Thousands joined during lockdown. → Thousands of users joined during lockdown, were seeking help for the first time.
2. Sentence-relative judgement: The data was shared for ad targeting. This directly contradicted the privacy promise. (one sentence, comma + which) →
3. Reduce both relatives: Clinicians who were employed by the platform assessed patients who were requesting stimulants. →
1. …lockdown, most of whom were seeking help for the first time.
2. The data was shared for ad targeting, which directly contradicted the privacy promise.
3. Clinicians employed by the platform assessed patients requesting stimulants.
Fill each gap with the relative structure the bracket requires, then read it aloud in a regulator's measured voice.
This was an industry (whose) central promise was confidentiality. The intake form, (non-defining: users completed it before treatment) , contained answers of the most sensitive kind. Those answers were shared with advertising platforms, (sentence-relative judgement: this was precisely what the policies promised would not happen) . The users affected — (quantifier: many were in crisis) — had no way of knowing. We note the standard (preposition + which) healthcare holds itself, and the companies (reduced: they were operating in this space) will now be held to it.
whose · which users completed before treatment · which was precisely what the policies promised would not happen · many of whom were in crisis · to which · operating in this space
Six structures, one paragraph — and note how the sentence-relative "which" lets a neutral document deliver its one moment of judgement.
Take the toolkit into the Speak tab — Task B's answer lives or dies on precise which-users, which-data distinctions.
A note on the record: figures are rounded; the FTC settlement involved no admission of wrongdoing; the prescribing investigation produced scrutiny and change rather than criminal findings; and the efficacy evidence for teletherapy is genuinely strong — this case's failures are failures of structure and incentive, not proof that remote care harms.
All of it out loud, against the clock. The register today is measured: this case involves real patients at their worst moments — argue it with the sobriety that earns the right to argue it at all.
Present the case to the board of a health insurer deciding whether to partner with digital-therapy providers. Structure it: the real problem and the real gains → the marketplace model and the pandemic detonation → the three reckonings: data, prescribing, financial gravity → what stabilised and on what evidence → the AI wave → the boundary question, and your partnering conditions.
Use at least four C1 relative structures and five key terms (telehealth, confidentiality, tracker, intake, controlled substance, settlement, engagement, pressured, vulnerability, lower barriers).
You are the new chief executive of a therapy platform, two years after your company's data settlement, at a health-policy conference. A clinician stands up: "Your industry took people's darkest disclosures and sold the fact of them to advertisers, and your defence was that everyone did it. Why should a single patient — or a single therapist — ever trust a platform again?"
Answer properly: concede the specific breach without the word "mistakes", separate the model from the abuse, state what is now structurally different — not promised, structural — and end on the patients the model still serves. Then compress it to ninety seconds.
You chair a working group writing the rules for digital mental-health services. Propose and defend three boundaries — one on data, one on prescribing, one on AI — each stated as a precise rule, each with its cost named honestly (every boundary prices someone out of access).
Prefer a different seat at the table? Argue the session as: