Research

Benchmarks for how answer engines cite, rank, and recommend.

Haynechi research turns AI answer behavior into the evidence teams need for planning, source strategy, and executive reporting.

Research Methodology

Benchmark evidence is inspectable, repeatable, and clearly labeled.

This sample methodology surface shows how Haynechi research moves from prompt collection to answer snapshots, citation normalization, human review, and board-ready artifacts while keeping sample rows distinct from published market data.

Scope benchmark

Run ProtocolSample

01 Prompt set

Buyer-stage clusters, competitor modifiers, region variants

scope locked

02 Engine run

ChatGPT, Gemini, Claude, Perplexity, Copilot, AI Overviews

captured

03 Citation parse

Owned, earned, partner, community, retail, and competitor sources

normalized

04 Answer review

Visibility, sentiment, missing entities, outdated claims, source drift

scored

05 Action link

Source plan, page brief, PR target, commerce fix, crawler check

assigned

Scoring ModelOperator reviewed

Visibility

brand presence, rank language, competitor displacement

Citation strength

authority, recency, source type, ownership path

Narrative risk

wrong claims, stale positioning, negative framing, omission

Movement readiness

clear owner, feasible action, measurable answer change

Quality GatesHonest proof

Sample-labeled rows

public UI separates demonstration data from real customer evidence

Source trace

each finding keeps prompt, answer, engine, region, URL, and capture timestamp

Repeatable run

same prompt family can be re-run to compare answer movement over time

Operator review

human review before reports become customer-facing or board-facing

Report ArtifactsReadout-ready

Benchmark appendix

methodology, prompt set, engines, regions, and sample limitations

Answer ledger export

snapshots, citations, deltas, sentiment, and capture metadata

Source opportunity map

pages and publishers most likely to influence future answers

Executive readout

what changed, what is risky, what to fund, and what to ignore

Benchmark Evidence Room

Make benchmark claims inspectable before they become board slides.

Research credibility comes from visible protocol, source taxonomy, confidence labels, rerun windows, and proof boundaries. This room turns prompt evidence into buyer-ready action while keeping sample rows distinct from published market data.

Report system

Evidence PacketsTraceable

Prompt protocol packet

Prompt family, buyer stage, region, competitor modifier, engine, capture window, and sample label stay attached.

protocol

Answer evidence packet

Answer text state, visibility, sentiment, missing entities, source set, and reviewer note are preserved together.

evidence

Citation trace packet

Citation URL, source class, ownership path, freshness, source strength, and correction opportunity remain inspectable.

trace

Movement caveat packet

Observed answer delta, inferred influence, crawler context, release marker, confidence label, and caveat stay separated.

caveat

Calibration RulesScored carefully

Sample before benchmark

Public examples must read as sample methodology unless a real dataset, scope, and publish approval exist.

Source taxonomy

Owned, earned, partner, retail, community, analyst, competitor, and generated sources stay separated.

Confidence labels

Observed, inferred, unknown, volatile, and outside-control states prevent benchmark language from overstating certainty.

Repeat windows

Prompt families rerun on agreed cadence so movement is compared against comparable engines, regions, and capture windows.

Readout PathEvidence to action

01 Lock scope

Define market, region, engines, prompt families, competitors, sample/customer-data state, and the first proof question.

02 Capture evidence

Run prompts, preserve answer snapshots, parse citations, classify sources, and attach capture metadata.

03 Calibrate findings

Separate visibility, citation strength, narrative risk, movement readiness, confidence, and caveats.

04 Route action

Turn findings into source plans, page briefs, PR targets, commerce fixes, crawler checks, and executive readout notes.

Reviewer LensesWho uses it

Buyer

Scope, first proof question, benchmark caveats, executive implication, and which finding deserves pilot work.

Operator

Prompt family, source opportunity, owner, action packet, evidence link, and expected answer movement.

Research

Protocol consistency, scoring model, source taxonomy, sample label, capture window, and rerun cadence.

Executive

Evidence of repeatable methodology, labeled sample scope, and clearly permissioned customer proof.

Evidence BoundariesLabeled scope

Labeled benchmark scope

Market-wide data, longitudinal depth, and customer adoption language stay tied to evidence state.

Causality with caveats

Answer movement, crawler activity, source changes, and referral quality carry caveats unless causality is proven.

Visible sample state

Sample methodology, conceptual examples, customer-approved evidence, and public benchmark data stay distinct.

Connected research output

Research points to an Answer Map, source plan, action backlog, Pilot Map, or proof question.

Guide Operating manual

Answer Engine Optimization Guide

A practical operating manual for prompts, citations, answer sentiment, and measurable movement.

Signal Inputs Prompt coveragedecision-stage prompts with tracked answer snapshotsCitation healthsource mix, freshness, and correction path

Workspace Outputs Prompt mapCitation auditAction backlog

Proof boundary

Research artifacts keep method, evidence scope, source class, and buyer-safe interpretation together.

View research

Report Benchmark report

AI Citations Trend Report

Benchmark the sources answer engines cite most often and the categories where trust shifts fastest.

Signal Inputs Cited domain classowned, earned, retail, partner, community, analystFreshness windowrecency of cited pages and crawl behavior

Workspace Outputs Source trend readoutCitation opportunity listCategory trust map

Proof boundary

Research artifacts keep method, evidence scope, source class, and buyer-safe interpretation together.

View research

Article Technical explainer

What Is llms.txt?

How technical teams can prepare crawler-facing guidance for AI systems without breaking web basics.

Signal Inputs Canonical coveragepriority pages included with stable descriptionsCrawler accessAI crawler events, response codes, and blocked paths

Workspace Outputs llms.txt draftCrawler-access checklistCanonical source map

Proof boundary

Research artifacts keep method, evidence scope, source class, and buyer-safe interpretation together.

View research

Comparison Comparison

Haynechi Vs Traditional SEO Tools

Where AI answer visibility diverges from rank tracking, keyword tools, and traffic dashboards.

Signal Inputs Search baselinerank, traffic, crawl, and conversion contextAnswer statevisibility, sentiment, citations, and competitor framing

Workspace Outputs Stack comparisonAEO gap mapBudget narrative

Proof boundary

Research artifacts keep method, evidence scope, source class, and buyer-safe interpretation together.

View research

White Paper White paper

Influence Orchestration

A governance model for connecting PR, content, commerce, and analytics around AI discovery.

Signal Inputs Claim consistencywhere public sources agree or conflictOwner coveragewhich source types have accountable teams

Workspace Outputs Governance modelOwner mapAnswer ledger

Proof boundary

Research artifacts keep method, evidence scope, source class, and buyer-safe interpretation together.

View research

Haynechi Index

Your category has an answer board.

See prompt, source, and narrative gaps before competitors make them default.

1 Category leader

2 Editorial incumbent

3 Your brand

4 Comparison challenger

5 Niche specialist