Octopus AI · Keyword intelligence · In build
A keyword layer built to explain, not just to list.
Octopus AI keeps its own keyword index: phrases and buyer questions with their market, intent, topic and source. It uses that index to match your pages to real demand and rank what to do first.
- Every record keeps its source and date
- Grouped into topics, not a flat list
- Mapped to the page that should answer it
Incoming · Sweet Crumbs
“eggless birthday cake hyderabad”
src · keyword dataset12 Oct 2026
Phrase
eggless birthday cake
enCommercial
Market
- India
- └ Telangana
- └ Hyderabad
Observation
Est. volume
320 / mo
12 Oct 2026
New observations are added, not overwritten. Data sources are being finalised for the pilot.
How the keyword layer works
Six steps from your site to a shortlist.
Octopus AI starts from what you actually sell, not from a generic keyword list, and every step can be traced back.
sweetcrumbs.example/cakes
We bake custom cakes for birthdays and weddings, including eggless options, and take corporate orders across Hyderabad.
Candidate buyer questions
Three kinds of record, kept apart.
A phrase, a local version of it and a measurement taken on a given date are different things. Octopus AI stores them separately so they are never confused.
The phrase
The query as typed or asked, its normalised form, its language and its search intent.
The market
The country, region or city the phrase applies to, because demand in one city says little about another.
The observation
A metric, such as estimated volume, recorded with where it came from and when. New observations are added, not overwritten.
What Octopus AI knows about each phrase.
Each record can hold the phrase, its language and location, estimated monthly demand, estimated difficulty, cost per click when available, search intent (informational, commercial, transactional or navigational), search result features when observed, the topic and parent phrase it belongs to, and its source and last update.
Fields with no reliable source stay empty. A blank is more useful than a number nobody can check.
- Phrase, language and location
- Volume, difficulty and CPC (estimated, sourced, dated)
- Intent and topic cluster
- Source and last updated
- phrase
- eggless birthday cake hyderabad
- language
- en
- location
- India › Telangana › Hyderabad
- intent
- commercial
- volume
- 320 / moest. · dataset · 12 Oct
- difficulty
- 34est. · dataset · 12 Oct
- cpc
- — (no source)
- serp features
- local pack, questionsobserved · 01 Nov
- topic
- Eggless cakes
- parent
- eggless cake
- updated
- 12 Oct 2026
A shortlist that shows its working.
Octopus AI extracts the services, products and places on your site, generates candidate questions, retrieves matching records, filters them by market, language, intent and relevance, groups them into topics, maps them to your pages and flags gaps.
Each opportunity then gets an Opportunity Score from 0 to 100, built from relevance, demand, feasibility, business impact and content gap. Omnishown Research shows the full method and the weights. The score is a guide to what to do first, not a ranking forecast.
- Retrieve, filter and group
- Map to existing pages
- Find gaps
- Score and explain
- 1Eggless cakes/cakes77
- · Matches a product you sell
- · Steady estimated demand
- · Page exists, needs an answer
- 2Corporate ordersGap69
- · Service named on your site
- · No page answers it yet
- · Higher order value
- 3Same-day custom cakes/order62
- · Matches your delivery promise
- · Competitive phrase
- · Small change to /order
- The score is a guide to what to do first, not a ranking forecast.
Retrieval first, then specialised scoring.
Octopus AI does not need a new foundation model. It combines fixed audit rules, retrieval from its own keyword index and your audit findings, and a language model that reasons over what was retrieved.
The plan is to start with retrieval, then tune the ranking and classification with properly authorised outcome data, then improve how recommendations are chosen using measured feedback from real use.
- Phase 1: retrieval over trusted data
- Phase 2: tuned scoring and classification
- Phase 3: improvement from measured feedback
Today's pipeline
- 01Fixed audit rules
- 02Retrieval · keyword index + your findings
- 03Language model · reasons over what was retrieved
No new foundation model needed.
The plan
- 1Retrieval over trusted dataFirst
- 2Tuned scoring and classificationPlanned
- 3Improvement from measured feedbackPlanned
What the keyword layer does not claim.
- Volumes, difficulty and cost per click are estimates. Each one says where it came from and when.
- A current ranking page is shown only when that data is actually available.
- The keyword data sources and their licensing are being finalised for the pilot.
Questions, answered plainly
Still unsure? Join the pilot and ask us directly. We'll answer before you connect anything.
Where does the keyword data come from?
The intended sources are licensed keyword datasets, authorised keyword or advertising APIs, Search Console data you choose to connect, permitted public search observations, and your own site's content. The final sources and licensing are being decided for the pilot. Every figure will show its source and date.
How is this different from Omnishown Research?
Research is the product you use: the ranked list and the Excel export. The keyword layer in Octopus AI is the index and the matching behind it.
Does it support my country and language?
The first target market and language are being decided with the pilot. Tell us yours when you join and we will say honestly whether it is covered.
Will it tell me exactly how many people search for a phrase?
No tool can know that exactly. Octopus AI shows an estimate with its source and date, and leaves the field blank when it has no reliable source.
Find out which questions matter for your business.
Join the pilot and tell us your market. We will say honestly what the keyword layer covers today.
- No card, no commitment
- Every fix arrives as a pull request you approve
- Planned pricing from ₹2,999 per site per month