Today's AI Specialist: The Language Analyst. The Agent Who Turns Every Blog Post Into a Lesson.

The blog is the front door of the platform. Most companies that run a blog stop there: the post is the deliverable, and the goal is reach. EnglishFluency.Online has a different problem. Every reader of the blog is also a learner. The post is not the deliverable. The lesson is.

The agent who turns the post into the lesson is the Language Analyst.

She runs after the Editorial Director approves the draft, after the SEO Specialist locks the metadata, before the post ships. Her output is the Learning tab that sits beneath every blog post on the live site, and the data that the AI coaches inside the platform pull from when a learner reaches for context.

She does not write the post. She turns the post into something a learner can study.

This is the build story of the Language Analyst. Why every post is a lesson, what she produces, and the design decision that decides whether the Learning tab is a useful supplement or a wasted column.

The problem the Language Analyst solves

A blog written for professionals who are also language learners has a dual audience inside every reader. The professional wants to absorb the argument. The learner wants to extract the language. The two readings happen at different speeds and serve different goals. Most blogs serve only the first. The professional reads, takes the point, leaves.

But the reader of an English-fluency blog is, by definition, also using the blog to improve their English. They are paying attention to the vocabulary they had to look up, the grammar structures they noticed, the phrases they want to remember. Without a structured surface to capture that second reading, the language learning is incidental and the reader has to do their own extraction.

The Language Analyst exists to do the extraction for them. Her job is to take the post, identify what is learnable in it, and surface that learnability in a tab the reader can open after they have finished the argument.

This turns the blog from a marketing surface into a learning surface, without changing the writing.

What the Language Analyst produces

Several artefacts per post, all written to a shared structure and rendered on the Learning tab.

The first three are translations of the body into Italian, French, and German. Not partial translations or summaries. Full renderings of the entire post, paragraph by paragraph. Each translation is a complete piece of writing in the target language; the reader can switch to Italian and read the entire post in Italian if they want to. The translations are not for replacement of the English: they are for scaffolding. A learner can read a paragraph in English, check the translation if a phrase is unclear, and return to the English without losing the thread.

The fourth artefact is the vocabulary list. A curated set of items selected from the body, each at the upper boundary of the post's likely reader's level. Every item carries the linguistic metadata a learner needs to study it on its own. The list is curated, not comprehensive. The agent is briefed to exclude generic vocabulary the reader already knows and highly specialised vocabulary the reader will not encounter again. What remains are the items with the highest carry-over to other professional English contexts.

The final artefact is the grammar plus comprehension layer. A handful of grammar notes, each tied to a specific construction in the post, with a worked example, an explanation, and a common mistake. Plus a set of comprehension questions, mixed across factual (did you follow), inferential (did you understand the logic), and application (can you use the framework in your own context).

The artefacts together turn the post into a structured lesson. The reader who only wants the argument reads the body and leaves. The reader who wants to learn from the post opens the Learning tab and works through the material at their own pace.

Decision one: the English is always primary

The single most important constraint on the agent is that the translations are scaffolding, not replacement.

The instinct, when you have full translations available, is to surface them as a language switcher: let the reader choose their language, read the whole post in their L1, never engage with the English at all. We tried that. The result was that L1 readers stopped reading the English entirely. The blog was no longer doing any English exposure work. The learning collapsed.

The fix was to keep the English as the primary rendering of the post on the page and put the translations behind the Learning tab as scaffolding. The reader who wants the L1 version has to ask for it, paragraph by paragraph, in the tab. The friction is deliberate. It keeps the reader in the English by default, and lets them check the L1 only when they need to.

This is the design decision that decides whether the blog produces actual exposure or only the feeling of accessibility.

Decision two: the vocabulary list is curated, not comprehensive

The second decision was the curation of the vocabulary list to a small, fixed ceiling.

A comprehensive vocabulary list would cover everything from the unusual to the routine. A comprehensive list looks more useful. It is not. The reader who sees a long vocabulary list at the bottom of a post will glance at it, conclude there is too much to learn, and not engage. A short, curated list is small enough to actually study.

The curation is the work. The agent has to decide which items are worth surfacing, which is a judgement call about what the post's reader is likely to encounter again. Generic professional vocabulary is excluded. Highly specialised vocabulary specific to one industry is excluded. What remains is the language that is at the upper edge of the reader's current level and will keep paying off in other professional contexts.

The hard ceiling forces the curation. Without the constraint, the list bloats and stops being useful.

Decision three: comprehension questions are mixed-type

The third decision is the structure of the comprehension layer.

Most language-learning content uses comprehension questions to check that the reader understood the literal text. Did the writer say X? Where did the writer mention Y? These are factual questions, and a comprehensive set of factual questions is what most platforms produce.

The Language Analyst is briefed to produce a mix of factual, inferential, and application questions. The factual questions check whether the reader followed the post. The inferential questions check whether they understood the post's logic beyond the literal text: why did the writer make this argument, what does this example imply. The application questions check whether they can use the post's framework in their own working life.

The application questions are the test of whether the post produced any change in the reader. A reader who can correctly answer a factual question has read the post. A reader who can answer an application question has internalised the post. The mixed-type structure surfaces the difference and lets the learner self-assess where they are.

TL;DR

The Language Analyst turns every blog post into a structured lesson. She runs after the Editorial Director approves the draft and produces a stack of artefacts: full Italian/French/German translations, a curated vocabulary list at CEFR level, a handful of grammar notes tied to specific constructions, and a mix of factual, inferential, and application comprehension questions. The single most important constraint: the translations are scaffolding, not replacement, and the English remains the primary rendering of the post on the page. The vocabulary list is kept short by design; comprehensive lists bloat and stop being useful. The comprehension questions are mixed-type because comprehension is not a single skill, and the application questions are the test of whether the post produced any change in the reader. The Language Analyst exists because every Behind the Build post should also be a lesson. The pedagogical discipline at work is the same one the eduQua dossier system holds Swiss continuing-education providers to. Applied per post, not per programme.

See how the Language Analyst was built and meet the rest of the team (/build)