In January 2025 I wrote about the four levels of speech described in ancient Indian texts: Vaikhari, Madhyama, Pashyanti and Para. They move from the gross to the subtle – from spoken words, to inner thought, to intuitive insight, to the silent source behind all expression.
Since then, one thought has stayed with me. Every day, our words are read not only by people but by machines – Google’s crawlers, Facebook and TikTok feeds, YouTube’s recommendation system, and now AI assistants such as ChatGPT and Gemini. What if this ancient framework could help us understand how those machines “listen” – and how we should write for them?
Here is the idea at the heart of this article:
Humans speak from the inside out – from Para to Vaikhari. Algorithms read from the outside in – from Vaikhari towards Para.
Good content is born at the deepest level and travels outwards into words. A machine starts with the words and tries to work its way back to the meaning, the intent and, finally, the trustworthiness of the source. The better your content holds together at every level, the further the machine can follow it – and the more confidently it will show it to people.
A quick reminder: the four levels of speech
- Vaikhari (expressed): audible, everyday speech – the words themselves.
- Madhyama (intermediate): inner, mental speech – the structure of thought behind the words.
- Pashyanti (seeing): intuitive perception – the flash of insight before words form.
- Para (transcendental): the unmanifest source – pure intention and consciousness, beneath everything.
Level 1 – Vaikhari: what the algorithm sees first
An algorithm’s first contact with your content is pure Vaikhari: the visible surface. Titles, headings, the words in your text, image file names and alt text, captions, hashtags, video subtitles, the link people click. Before any machine can understand you, it has to be able to read you.
AI systems literally break your text into small pieces (tokens) before anything else happens. If the words are vague, hidden inside images, or buried in clever wordplay, the machine starts from a weak position.
How to write for Vaikhari:
- Use the words your audience actually uses – including in Vietnamese and English if you serve both.
- One clear title and clear headings that describe what each section is about.
- Put important text in real text, not only inside images or graphics. Add alt text and subtitles.
- Write short, clean sentences. Clarity is a form of respect – for people and machines.
The trap: keyword stuffing. It is Vaikhari shouting with nothing behind it. Like chanting a mantra loudly while the mind is elsewhere, it makes noise but carries no power – and modern algorithms recognise it.
Level 2 – Madhyama: how the algorithm understands meaning
Madhyama is inner speech – the structure of thought behind the words. For a machine, this is the level of meaning: which topics you cover, how ideas connect, what entities you mention (people, places, products, services) and how they relate.
Modern search and AI systems turn text into mathematical representations of meaning, so that “dịch vụ Google Maps” and “Google Business Profile management” can be recognised as related ideas even when the words differ. This is the machine’s version of Madhyama: it is no longer just hearing the words, it is following the thought.
How to write for Madhyama:
- Give every piece one main idea, and let each section develop it logically.
- Answer the natural follow-up questions a reader would ask – this is how inner dialogue works.
- Name things precisely: your business, location, services, products, people.
- Link related content together on your website so the machine can see how your ideas connect.
- Use structured data (schema) where it fits – it is a direct way to tell machines what your content means.
Level 3 – Pashyanti: how the algorithm senses intent and value
Pashyanti is “seeing” – the moment of insight when something simply clicks. In content, this happens in the reader: “Yes, this is exactly what I needed.”
Algorithms cannot feel that moment, but they watch for its traces. Did people stay and read? Did they save, share, comment or come back? Did the content match what they were really looking for when they searched or scrolled? These signals are the machine’s way of inferring Pashyanti – whether your words created real understanding.
How to write for Pashyanti:
- Start from the reader’s real intent, not from what you want to sell. What problem are they trying to solve right now?
- Give the key insight early. Don’t make people dig for the value.
- Offer something original: your own example, observation, experience or framework – something they cannot get from a hundred similar posts.
- Write so that a reader would want to send it to someone else. Sharing is Pashyanti made visible.
Level 4 – Para: the source that algorithms try to trust
Para is the unmanifest source – silent, invisible, yet present behind every word. In content, Para is who you really are and why you are speaking: your genuine experience, expertise, intention and integrity.
No algorithm can see Para directly. But increasingly, search engines and AI systems try to judge it through its reflections: Is there a real, identifiable author or business behind this content? Is it consistent over time? Do others trust it – through reviews, mentions, links and reputation? Google’s own guidance speaks about experience, expertise, authoritativeness and trustworthiness. That is, in a sense, an attempt to measure the source.
How to write from Para:
- Write about what you genuinely know and have done. Experience cannot be faked for long.
- Show who is speaking: a real author name, a real business, real contact details, a verified Google Maps profile.
- Be consistent – the same voice, values and information across your website, Google, Facebook, Zalo and LinkedIn.
- Let your intention be service, not manipulation. As with mantra practice, the purpose behind the words shapes their effect.
Putting it together: a four-level content check
Before you publish anything – a blog post, a Facebook post, a product page or a video – ask four questions:
- Vaikhari: Are the words clear, readable and in the language my audience uses?
- Madhyama: Is there one clear idea, logically developed and connected to my other content?
- Pashyanti: Will the reader feel “this is exactly what I needed” – and want to share it?
- Para: Does this come from real experience, and is it clear who is speaking and why?
Most content today is written only at the first level – words arranged for a machine. The content that lasts is written from the fourth level and made clear at every level on its way out.
Why this matters even more with AI
As AI assistants increasingly answer questions directly, they choose a small number of sources to rely on. They favour content that is clear on the surface, coherent in meaning, genuinely useful and coming from a trustworthy source. In other words: content that is aligned on all four levels.
The irony is beautiful. The more advanced our machines become, the more they reward something very old – speech that is true from its source to its sound.
How DMG can help
At DMG, we help businesses in Vietnam create content that works at every level: clear on the surface, well structured for search and AI, useful to real customers and rooted in genuine expertise. From content strategy and SEO to Google Maps (Google Business Profile) optimisation, we make sure both people and algorithms understand what your business truly stands for.
Want to see how your website reads to machines today? Start with our Free Website Check, or message us on WhatsApp.
This article builds on my earlier piece, Exploring the Four Levels of Speech, on LinkedIn. I welcome your thoughts – especially from those who practise mantra and those who work with AI every day.
