Skip to content

Typewise AI Discoverability Review

5 min read

Preliminary observations & hypotheses

Prepared by
Anindo Neel Dutta

Approximately 4 hours of exploratory research

Last updated on
9th July, 2026

"This isn't an exhaustive audit. It's a demonstration of how I approach unfamiliar problems."

Introduction

I became interested in Typewise after noticing an interesting pattern while working on one of my own products, DocPilot.

Despite very little active marketing over the past few months, DocPilot continued receiving citations from AI systems. That observation made me curious about how modern LLMs discover, retrieve, and recommend software products.

When I started researching Typewise, I expected to find a company that was consistently recommended across common enterprise AI customer support queries.

Instead, I found a noticeable gap between the quality of the product and how frequently it was recommended by AI systems.

This document summarizes my initial observations, hypotheses, and what I’d prioritize if I joined the team.

Observations

Observation 1

Product quality doesn't match AI visibility

I tested a series of common enterprise customer support discovery queries across ChatGPT, Claude, Gemini, and Perplexity.

Examples included:

  • Best enterprise AI customer support platform
  • Best AI support agent software
  • Enterprise AI helpdesk
  • AI customer support software
  • AI customer service platform
  • Zendesk alternatives

Across these queries, competitors such as Zendesk, Intercom, Ada, Salesforce, Kore.ai, NICE, and Botpress appeared consistently across multiple AI systems.

In contrast, I rarely saw Typewise recommended—even for queries where I expected it to be highly relevant. This pattern was consistent enough that it immediately stood out to me.

Frequently surfaced

ZendeskIntercomAdaSalesforceKore.aiNICEBotpress

Rarely surfaced

Typewise

Observation 2

This doesn't appear to be an engineering problem

From what I observed, Typewise already has many of the qualities I’d expect from a leading enterprise AI company.

  • Strong visual identity
  • Fast, polished website
  • Enterprise positioning
  • Strong customer credibility
  • Industry recognition

These don’t look like the bottleneck.

Instead, I think there’s an opportunity to improve how AI systems understand, categorize, and retrieve information about Typewise.

Hypothesis

My current hypothesis is that discoverability is limited more by information architecture and technical messaging than by product quality.

A few examples stood out to me.

The homepage communicates customer outcomes very well, but developer capabilities such as APIs, SDKs, integrations, and technical workflows require more exploration than I expected.

Because LLMs rely heavily on explicit, structured technical information, I think these capabilities deserve much greater prominence throughout the website and documentation.

Similarly, I’d expect more implementation-focused content such as:

  • API guides
  • SDK documentation
  • Integration tutorials
  • Architecture documentation
  • Technical comparison pages
  • Developer use cases

This type of content is useful for developers, but it’s also exactly the sort of information AI systems can reliably retrieve and reference.

Current state

Strong product and brand presence, but technical capabilities and implementation detail are harder for AI systems to surface.

Opportunity

Elevate APIs, SDKs, integrations, and architecture content across the site and documentation.

Expected outcome

Clearer categorization in AI responses and more frequent recommendations for relevant enterprise queries.

Why this matters

LLMs don’t recommend products simply because they’re the best.

They recommend products because they repeatedly encounter clear, structured, and consistent information describing:

  • What a product is
  • Who it’s for
  • What problems it solves
  • When it should be recommended

The easier this information is to retrieve and reinforce, the more frequently a product tends to appear during AI-assisted product discovery.

A small experiment from my own work

The observation that sparked my interest in this space came from DocPilot.

Although I hadn’t meaningfully worked on the product for nearly a year, Bing Webmaster Tools began reporting sustained AI citations around March.

185

AI Citations

118

Cited Pages

76 / 90

Active Days

11

Peak Daily Citations

Bing Webmaster Tools AI Performance report showing citation trends for DocPilot

Source: Bing Webmaster Tools — AI Performance Report

I don’t present this as proof of success. Rather, it sparked my curiosity about why some products consistently appear in AI recommendations while others don’t.

That curiosity is what eventually led me to analyze companies like Typewise.

What I'd do in my first 30 days

  1. Week 1

    Establish a baseline

    • Audit AI mentions across major LLMs
    • Benchmark competitors
    • Review technical documentation
    • Evaluate structured data and information architecture
  2. Week 2

    Strengthen technical positioning

    • Prioritize clearer messaging around APIs
    • Surface SDKs and integrations
    • Document MCP where applicable
    • Clarify developer workflows
  3. Week 3

    Build AI-readable content

    • Integration guides
    • Technical implementation tutorials
    • Comparison pages
    • Architecture documentation
    • Enterprise deployment examples
  4. Week 4

    Build a measurement system

    • Monitor AI citations and LLM visibility
    • Track query coverage and competitor movement
    • Iterate continuously based on observations

Methodology

Final thoughts

Thanks for taking the time to read this.

If you’d like to discuss any of these ideas, or tell me where you think I’m wrong, I’d genuinely enjoy the conversation.

Anindo Neel Dutta