Keyword research for typed text-only queries are hitting a wall. People ask questions out loud to voice assistants, take a photo and ask “what is this and where can I get one” or have a back-and-forth conversation with ChatGPT or Gemini that refines an answer over five or six turns. None of those behaviors are cleanly captured by a traditional keyword list sorted by search volume.
This change means that for brands investing in AI SEO services in Indore, keyword research itself needs to change shape. Now it’s no longer just about finding the right words, it’s about mapping the underlying intent behind spoken questions, image-based searches and multi-turn conversations and then building content that satisfies that intent, irrespective of what format the question came in.
In this post I’ll walk through how intent-based keyword research works in a multimodal, conversational search world, and how it fits into a broader generative engine optimization strategy.
The Pitfalls of Traditional Keyword Research
Traditional keyword research begins with a seed term, pulls volume and difficulty data, and builds a list of variations. That approach assumes:
Queries are typed rather than spoken – Each query is a separate search, not a conversation – Intent is mostly inferred from the words themselves.
Multimodal and conversational search violates all three of these assumptions. A voice query such as “where can I get my phone screen fixed near me right now” has a different phrasing, urgency and context than the typed equivalent “phone repair shop”. There’s no way at all to express a broken screen image search – the intent must be inferred from visual content plus a short spoken or typed follow-up. And a conversation with an AI assistant often starts broad and narrows over several turns, meaning the ‘real’ query a business needs to rank for is burying three messages into the conversation.
The keyword tools built for the old model just don’t catch this behavior.
The 3 Levels of Modern Search Intent
- Textual Intent – the traditional layer. What will people type into a search box? This is still important and still needs data on volume and competition.
- Conversational intent — the same need expressed in a multi-turn exchange with an AI assistant. This includes follow up questions, clarifiers and the natural language phrasing people use when talking vs typing.
- Multimodal intent — how a single need is expressed in images, voice, or a combination of both. A user might take a picture of a product and ask, “is this worth buying”. He wants an answer in comparison, not a definition.
Good keyword research today maps all three layers for a given topic, rather than stopping at the first.
