TextCortex AI Review
TextCortex is the best-run company in this category and it still fails the simplest test there is: whether its own documents agree with each other about what happens to your data.
Scored on the six-dimension BrokenCtrl methodology, set independently of the affiliate link. How scoring works.
| Dimension | Basis | Score |
|---|---|---|
| Transparency | Pricing FAQ, privacy policy and opt-out form give three different answers on training. | 4 / 10 |
| Data Privacy | Free AND paid users trained on by default; only Enterprise contractually exempt. | 3 / 10 |
| Safety Architecture | Product itself competent; no documented harm events. | 5 / 10 |
| Corporate Conduct | Best-run company in the cohort, which makes the contradictions a choice. | 4 / 10 |
| Bias Mitigation | No published bias evaluation. | 5 / 10 |
| Regulatory Alignment | DPO is also managing director — a GDPR Article 38(6) conflict of interest. | 4 / 10 |
| Total | 25 / 60 |
The pricing FAQ asks “Will TextCortex use our data to train AI models?” and answers: “Absolutely not. None of your data will ever be used for model training or fine-tuning.”
The privacy policy, same domain, says: “we may utilize the data you provide to enhance and refine our AI models.”
And there is a live opt-out form at tell.textcortex.com/opt-out which reads, verbatim: “when you use TextCortex, we may use the data you provide us to improve our models.”
You cannot offer an opt-out from something that never happens. The form’s existence is self-refuting proof that the “absolutely not” is false. The opt-out is also prospective only — it does not reach what has already been ingested.
The usual assumption — free tier gets trained on, paid tier is safe — does not hold here. Free and paid Premium users are trained on by default, opt-out only. Only Enterprise is contractually exempt. The $29.99 customer is on the training side of the line and is unlikely to know it.
The privacy policy names Ceyhun Derinbogaz as Data Protection Officer. The imprint names the same person as managing director. Under GDPR Article 38(6), a DPO must not hold a role that determines the purposes and means of processing — which is what a managing director does by definition. This is a textbook independence conflict, and it is visible from a cross-read of two of the company’s own pages.
Give credit where it is due: TextCortex has a real trust portal, and the URL structure indicates SOC 2 Type II. That puts it far ahead of every competitor in this cohort. But the portal renders nothing to a non-browser client, so the auditor, the audit period and the report-access terms could not be established. The DPA is sales-gated. There is no sub-processor list at all.
“Hosted exclusively in the EU” is contradicted by TextCortex’s own model table, which lists Gemini, Kimi and Grok in US regions and DeepSeek in GB.
TextCortex marketed a proprietary stack — NeoCortex, Sophos, Velox, Alta — presented in its model picker as a peer of GPT-4 and Claude. Its current help-centre model list enumerates every selectable model: Alibaba, Anthropic, DeepSeek, Google, Moonshot, OpenAI, Z.AI, xAI.
Not one TextCortex-built model appears on it. No model card, no paper, no benchmark was ever published for any of them. The homepage now honestly says “model-agnostic” — but the AI content policy still governs “NeoCortex” in the present tense. It is an ethics policy for a model that no longer exists in the product.
Free, Premium at $29.99 per user per month, and custom Enterprise. Each seat carries a token wallet worth roughly half the licence fee; exhaust it and you silently fall back to weaker models. Overage is billed at base LLM cost plus a platform markup.
Textly / Linguix — Same category, same disclosure disease, worse paperwork.
Creaitor — The other European writer whose privacy policy forgets its AI processor.
Does TextCortex train on my data?
Its pricing FAQ says never. Its privacy policy says it may. A live opt-out form exists, which only makes sense if training happens. Free and Premium users are opted in by default; only Enterprise is exempt.
Is TextCortex SOC 2 certified?
It claims SOC 2 and ISO 27001, and it has a real trust portal — better than any competitor here. The auditor, audit period and report terms could not be independently established.
Does TextCortex have its own AI model?
Not any more. Its current model list contains only third-party models — OpenAI, Anthropic, Google, xAI and others. None of its previously marketed in-house models appear.
Is TextCortex hosted in the EU?
It says so, but its own model table lists models running in US and GB regions.