Get the newsletter
Google · AI / ML · Filed Nov 20, 2025 · USPTO published Jun 11, 2026 · Jun 11 release

Google Patents a System That Tells You Which AI Prompts Will Actually Work

Getting good results from an AI chatbot often comes down to how you phrase your request — but most people have no idea whether their wording is good or bad until they see a mediocre answer. Google is patenting a system that shows you, before you even hit send, which suggested prompts are the best fit for what you've already typed.

By · Updated · 3 min read

Google Patent: AI Prompt Suggestions With Compatibility Scores, figure from US 2026/0161268 A1
Figure from the official USPTO publication.
Publication number
US 2026/0161268 A1(US20260161268A1)
Applicant
GOOGLE LLC
Filed · Published
Nov 20, 2025 · Jun 11, 2026
Inventors
Ajay Prasad, Ramprasad Sedouram, Gulmohar Khan
Status when we published
Waiting for an examiner
Document
20 claims
US classification
715/256
AI assistant & agent

What Google's prompt-suggestion feature actually does

Imagine you start typing a request into an AI assistant — something like "summarize this article" — and the app immediately shows you a list of suggested follow-up phrases or ways to complete your prompt. That already exists in some tools. What Google's patent adds is a visual indicator on each suggestion showing how well it matches what you've already written.

Think of it like a Wi-Fi signal bar for your prompt: some suggestions are a strong match for your input, others are a weaker fit. The patent says this compatibility signal can also hint at whether a suggestion will use more or fewer of the AI's processing resources — meaning some phrasings are cheaper and faster to run than others.

You can also interact with a suggestion (say, by tapping it on a touchscreen) to reveal that compatibility detail. Once you pick one, the app combines your original input with the chosen suggestion and sends the whole thing to the AI to generate your answer.

How the compatibility indicator scores each suggestion

The patent describes a generative AI application — think a chatbot or AI writing tool — that does three things in sequence.

First, as a user types an initial input, the system uses one or more generative models (the same kind of large language model that powers tools like Gemini or ChatGPT) to generate a list of prompt suggestions: pre-written phrases or completions that, when combined with the user's input, should produce better AI output.

Second, each suggestion is rendered as a selectable GUI element (a button or chip on screen) that carries a compatibility indicator — a visual feature signaling how well that suggestion aligns with what the user already typed. The patent notes this indicator can reflect two things:

  • Whether the suggestion will produce a more accurate AI output
  • Whether it will consume fewer tokens (tokens are the units of text an AI model processes; fewer tokens generally means faster and cheaper responses)

Third, when the user selects a suggestion, the app combines it with the original input and feeds the combined prompt to the AI model to generate the final content. The compatibility detail can also appear in response to an interaction gesture — like a tap or hover — rather than being shown by default.

What this means for everyday AI chatbot users

For most people, prompting an AI is still guesswork. You type something, get a so-so answer, rephrase, try again. A built-in signal that says "this phrasing is a strong match for what you've typed" could meaningfully cut down that trial-and-error loop — especially for users who are new to AI tools and don't yet have intuitions about what makes a good prompt.

For Google, this fits neatly into its Gemini ecosystem, where the company is clearly trying to make AI assistance feel less intimidating and more guided. The token-efficiency angle is also interesting: if Google can steer users toward prompts that cost less to process, that's a real operational saving at scale across millions of queries.

Source: US 2026/0161268 A1 at the USPTO (PDF). Full patent text and figures are in the official publication.

Get our take in your Top Stories

Liked this breakdown? Add Patentlyze as a preferred source on Google, and our plain-English take shows up more often in your Top Stories the next time Google patent news breaks.

Add Patentlyze on Google One click.
There are more where this came from

We read every patent application Big Tech publishes and send you the ones worth knowing. Plain English, free, every week.

More from Google

All 1,079 Google patents →

More in AI assistant & agent

All 413 AI assistant & agent patents →