TypeSafe AI’s Jev can classify whether two proposed blog topics serve distinct search intents or risk becoming competing pages. The application supplies two structured briefs as state, and Jev answers a Choice question with one option from a defined overlap taxonomy. It does not search Google, calculate SERP overlap, propose a new angle, or explain the decision in prose.
That distinction is important for SEO planning. Code should first compare assigned keywords, existing URLs, and observed search results. Jev can then evaluate the less mechanical question: do these two briefs promise meaningfully different value to the same reader? The returned choice, probabilities, and confidence help an editor decide whether to approve both topics, refine one brief, consolidate them, or gather more evidence.
What Blog Topic Overlap Means
Two titles can share words without competing. “What Is Keyword Cannibalization?” and “How to Diagnose Keyword Cannibalization in Search Console” are related, but one answers a definition query and the other supports a diagnostic task.
The opposite is also true. Two titles can use different wording and still promise the same outcome. “Build a Better Content Calendar” and “Plan Monthly Blog Production” may target the same audience, problem, format, and next action.
Content overlap becomes an SEO risk when pages share most of these elements:
- primary reader;
- search intent;
- core question;
- expected answer format;
- promised outcome;
- conversion goal;
- set of ranking competitors.
This is different from literal duplicate content. The copy can be completely different while the pages still compete for the same need. Before creating another URL, compare what a searcher expects each page to do.
Screpy’s guide to planning topic clusters explains how related pages can support one another when their jobs are clearly separated. The Jev workflow in this guide focuses on the narrower approval decision between two briefs.
What TypeSafe AI’s Jev Contributes
Jev is TypeSafe AI’s first System One Model. It evaluates typed questions against supplied state and returns structured decisions. For topic overlap, Choice is appropriate because the answer must come from a fixed set of editorial outcomes.
Jev does not produce free-form reasoning. It cannot:
- browse the live SERP;
- discover existing CMS pages;
- calculate URL or ranking overlap;
- write a revised title;
- invent a differentiating angle;
- recommend a redirect in prose;
- ask an editor follow-up questions.
Your application gathers those facts before the call. After the call, code maps the typed result to an editorial route. A writer or strategist handles creative changes.
Collect Evidence Before Comparing Briefs
A topic comparison should use more than titles. Build the state from normalized brief data:
| Field | Why it matters |
|---|---|
| Working title | Shows framing, but not the complete intent |
| Primary query | Identifies the intended demand |
| Audience | Separates beginner, specialist, buyer, and developer needs |
| Main question | States the problem the page must answer |
| Format | Distinguishes definition, tutorial, comparison, template, or case study |
| Promised outcome | Describes what the reader can do afterward |
| Scope boundaries | Prevents related pages from covering the same material |
| Conversion goal | Separates education from product evaluation |
| Existing URL | Reveals whether the topic already has a canonical home |
| SERP overlap | Provides observed evidence that Google treats queries similarly |
SERP overlap should be calculated outside Jev. Run both target queries in the same market and device, normalize the ranking URLs, and calculate how many top results appear for both. Keep the raw number in state as evidence; do not ask Jev to pretend it performed the search.
The CMS comparison should also be deterministic. Search published titles, slugs, assigned queries, and descriptions before sending a new pair for judgment. Screpy’s guide to finding competitor content gaps can inform topic discovery, but a gap is not automatically a reason to create a separate page.
Define a Clear Overlap Taxonomy
Use option descriptions that lead to distinct editorial actions:
- distinct: different audience need or outcome; both pages can stand alone;
- related_but_separate: connected topics with clear boundaries and useful cross-linking;
- partial_overlap: meaningful shared coverage; one or both briefs need tighter scope;
- same_intent_duplicate: substantially the same reader, question, and outcome;
- insufficient_context: the supplied briefs do not support a reliable decision.
The insufficient-context option prevents false precision. If neither brief defines audience or outcome, the correct action is to improve the inputs before approving another URL.
Avoid options such as “good” or “bad.” They do not describe the relationship between pages and cannot map cleanly to an editorial workflow.
Example Jev Choice Request
The request below compares two briefs. Jev sees the state and the human-readable option criteria. It does not receive a generic chatbot prompt.
{
"model": "jev-latest",
"state": {
"topic_a": {
"title": "What Is Keyword Cannibalization?",
"audience": "SEO beginners",
"primary_query": "what is keyword cannibalization",
"main_question": "What does keyword cannibalization mean?",
"format": "definition guide",
"promised_outcome": "Recognize the concept and common causes"
},
"topic_b": {
"title": "How to Find Keyword Cannibalization in Search Console",
"audience": "SEO practitioners",
"primary_query": "find keyword cannibalization in search console",
"main_question": "Which queries and URLs are competing?",
"format": "step-by-step diagnostic workflow",
"promised_outcome": "Build a list of competing query-page pairs"
},
"shared_top_10_urls": 3
},
"questions": {
"topic_relationship": {
"type": "choice",
"instructions": "How should these two proposed pages be classified based on audience, intent, format, and promised outcome?",
"criteria": {
"distinct": "The pages solve different problems and do not need a shared editorial structure.",
"related_but_separate": "The pages are connected but each has a clear standalone intent and outcome.",
"partial_overlap": "The pages share substantial coverage and need stronger scope boundaries.",
"same_intent_duplicate": "The pages serve substantially the same reader, question, format, and outcome.",
"insufficient_context": "The supplied briefs do not contain enough information to classify reliably."
}
}
}
}
The relationship is more likely related-but-separate because one page explains a concept and the other performs a specific diagnostic task. That expectation should still be tested against reviewed examples rather than assumed.
Read the Choice Output
The response below is illustrative. The numbers demonstrate the response shape and are not measured accuracy.
{
"model": "jev-1.13.0",
"answers": {
"topic_relationship": {
"type": "choice",
"choice": "related_but_separate",
"confidence": 0.74,
"probabilities": {
"distinct": 0.05,
"related_but_separate": 0.78,
"partial_overlap": 0.14,
"same_intent_duplicate": 0.01,
"insufficient_context": 0.02
}
}
}
}
The selected option is the highest-probability answer. Confidence describes the concentration of the distribution, not the correctness of the SEO decision. Here, partial_overlap still has some probability, which tells the editor to verify that the definition article does not absorb the full diagnostic workflow.
Keep the complete distribution. A result split between related_but_separate and partial_overlap should not be reduced to a green approval badge.
Route the Decision in Code
Map each typed result to a clear action:
| Choice | Editorial route |
|---|---|
| distinct | Approve both briefs independently |
| related_but_separate | Approve with explicit boundaries and cross-links |
| partial_overlap | Return both briefs for scope revision |
| same_intent_duplicate | Pause the new URL and review consolidation |
| insufficient_context | Request missing brief fields |
| Low confidence | Send to an SEO strategist |
Jev does not write the revised angle. The review screen should show the two briefs, deterministic signals, and typed distribution. An editor can then change the audience, outcome, format, or scope until the pages have distinct jobs.
For same-intent cases, inspect existing performance before choosing consolidation, redirect, or cancellation. Rankings, backlinks, conversions, freshness, and business importance are facts that belong in the final human decision.
Prevent Cannibalization Before Drafting
The cheapest overlap problem is the one stopped before writing begins. Add this check to topic approval:
- Normalize the proposed query and topic.
- Search the CMS for related published and planned URLs.
- Compare current keyword assignments.
- Measure live SERP overlap where query data is available.
- Package two complete briefs as state.
- Ask one bounded Choice question.
- Route uncertain or overlapping results to review.
- Record the approved page purpose and exclusions.
- Recheck the pair if the brief changes materially.
This workflow creates a durable content map. The approved intent, audience, and promised outcome can later be used for internal linking, refresh decisions, and performance analysis.
Google’s guidance on people-first content supports creating pages for a real audience and purpose rather than expanding content merely to capture more query variations.
Keep This Page Distinct From General Topic-Cluster Advice
A topic-cluster guide answers how to organize a pillar and supporting pages across a subject. This Jev guide answers a different query: how to classify the relationship between two proposed briefs with a typed decision.
It should not become another general article about keyword research, editorial calendars, or pillar pages. Its unique value is the operational contract:
- normalized brief fields in;
- one defined Choice taxonomy;
- probabilities and confidence out;
- code-based routing;
- human scope decision.
For another real SEO classification pattern, see Screpy’s article on classifying search intent with TypeSafe AI’s Jev.
Frequently Asked Questions
Can Jev search the CMS for overlapping posts?
No. Your application must retrieve possible matches from the CMS and include the relevant briefs in state.
Can Jev calculate SERP overlap?
No. Collect rankings with an appropriate search-data source and calculate shared URLs in code. Jev can use the resulting signal as supplied context.
Can Jev suggest a new article angle?
No. Jev returns typed decisions rather than generated text. An editor or separate generative model can draft a new angle after the classification.
Does a related topic always cause keyword cannibalization?
No. Related pages can support one another when they satisfy different intents and have clear scope boundaries. The risk increases when audience, question, format, and outcome substantially match.
Jev is useful here because it turns a fuzzy editorial question into a stable classification. The surrounding SEO system still supplies evidence, and the editor still decides what deserves a separate URL.