Measure whether your content is accessible, structured, attributable, and ready to cite—then compare its signals across ten major AI model families.
◎
Built-in analysis. Independent recommendations. Transparent estimates. No API keys · No model accounts · No technical setup
MODEL FAMILIES10
EVIDENCE LAYERS03 measured · built-in · modelled
YOUR INDEXHALO WORKSPACE
Reports, monitoring, and change history.
Sign in to keep reports together. Professional accounts can crawl entire websites, schedule recurring checks, compare snapshots, and track live citation observations where configured.
Your workspace will appear here.
Report history, monitored websites, last crawl, next scheduled run, score movement, and citation changes are shown in one place.
Unlock all ten models, every extracted claim, source gaps, and the complete action plan.
FROM CONTENT TO EVIDENCE
AI visibility has signals. Make them useful.
IndexHalo turns observable content and crawl signals into decisions your team can act on—while clearly separating direct findings, optional live checks, and comparative estimates.
01TRACE
Know what is retrievable.
See the claims, headings, dates, source links, metadata, and crawler signals that IndexHalo can observe directly.
02VERIFY
DIRECT CRAWLMEASUREDSIGNAL POOLBUILT INMODEL PROFILESESTIMATED
Know what the score means.
Every finding is labelled by evidence type, so your team can defend the decisions it makes.
03IMPROVE
PRIORITY MATRIX01
Act on the next best move.
Recommendations are ordered by impact and effort so teams know what to improve first—without invented uplift claims.
MODEL COVERAGE
One report. Three ecosystems.
Coverage follows how AI is distributed—from global assistants to specialised APIs and regional model families.
Reach figures are product-planning estimates and are labelled as such in the report.
TIER 01
Billion-user scale
Ubiquitous ecosystem integration
Meta AIOpenAIGoogle Gemini
EST. REACH~950M–1.2B
TIER 02
Enterprise + specialist
High-value professional ecosystems
Microsoft CopilotClaudeGrok
EST. REACH~220M–420M
TIER 03
Developer + emerging
API, open-weight, and regional scale
DeepSeekQwenKimiMistral
EST. REACH~25M–100M
A REPORT YOU CAN DEFEND
From raw content to usable evidence.
IndexHalo follows a staged evidence process. It observes what is publicly available, maps citation-relevant signals, compares those signals across model profiles, and turns gaps into a ranked action and investment plan.
IndexHalo reads declared sitemaps, follows same-domain links, respects crawler rules, rejects private-network targets, and records every analysed, blocked, redirected, or failed page.
INPUT02
EXTRACT
Build a page-level evidence inventory.
Each page is checked for headings, claims, dates, authorship, canonical metadata, structured data, source links, and crawler access. Scanned PDFs and supported images pass through OCR.
MEASURED03
POOL
Group the signals.
The observed findings are organised into six understandable categories: retrievability, authority, specificity, structure, freshness, and transparency.
MEASURED04
COMPARE
Read the signals through ten profiles.
The same evidence is compared across ten model-family profiles. These are disclosed estimates based on the content signals—not access to any vendor’s private ranking system.
MODELLED05
OBSERVE
Run repeatable citation checks.
When IndexHalo’s server-side provider connections are available, it stores the exact query, timestamp, returned citations, target position, and competing domains. Customers never supply API keys.
VERIFIED06
PRIORITISE
Fund the work with the strongest case.
Every recommendation states the affected pages, owner, current condition, target, effort and recrawl test. Imported demand and business value add conservative, base and upside scenarios without being disguised as guaranteed outcomes.
ACTIONABLE07
MONITOR
Re-run and compare over time.
Daily, weekly, or monthly snapshots feed a longitudinal citation observatory that separates comparable runs, calculates provider and source movement, detects volatility, and identifies the domains replacing lost visibility.
SCHEDULED08
ALERT
Escalate material changes while they still matter.
Deterministic alert rules connect crawl, citation, competitor, crawler-log, performance and workflow evidence to severity, owner, response target, exact action and automatic resolution.
OPERATIONAL09
EXPERIMENT
Prove whether the intervention moved the evidence.
IndexHalo freezes the target page, question set, provider set, competitor set, demand window and crawl snapshot before work begins. Later observations are evaluated only when those boundaries remain comparable.
CONTROLLED
WHAT THE REPORT CAN PROVE
What IndexHalo fetched, extracted, measured, and observed in timestamped provider responses—including the exact citation links returned in those checks.
WHAT NO INDEPENDENT TOOL CAN PROVE
Undisclosed private training data or guaranteed future citations. IndexHalo never disguises estimates as direct vendor evidence.
This worked example follows one coherent business problem from crawl evidence to a 30-day implementation plan. Every number is illustrative, but every field, calculation and recommendation represents the practical detail a Professional report is designed to provide.
FICTIONAL CLIENT · ILLUSTRATIVE DATA · REAL REPORT STRUCTURE
SAMPLE PROFESSIONAL REPORT · HELIOFORGE RESEARCH30-DAY GEO GROWTH BRIEF
CLIENTHelioForge Research
Fictional European publisher of commercial-solar benchmarks for finance, property and procurement teams.
COMMERCIAL OBJECTIVEOwn the answer for solar-payback decisions
Increase qualified visibility when buyers compare payback periods, ownership models, incentives and PPA pricing.
MEASUREMENT SCOPE24 buyer questions × 3 providers
72 timestamped observations, one 128-URL crawl, a seven-day crawler-log window and a comparison against returned sources.
EXECUTIVE DECISIONRepair and repackage 12 high-intent pages before commissioning more general articles.
The site already contains differentiated data. Its immediate problem is access, source attribution and answer packaging—not a lack of subject matter.
CRITICAL · WEB OPERATIONS7 crawler-requested URLs return 403
Respond within 4 hours · resolve when later 2xx events arrive for every URL.
HIGH · GEO + EDITORIALCountry-payback citation coverage is 2/18
Strengthen the owning page · verify with the same 18 questions and providers.
OPPORTUNITY · CONTENT STRATEGYGermany versus UK has no credible page owner
Publish the prescribed comparison asset · resolve at ≥60% relevance.
ACCESS86121/128 URLs retrievableEVIDENCE4866/112 claims lack a nearby sourceANSWERABILITY5831 pages lack concise answer blocksSTRUCTURE74103 pages have a clear hierarchyFRESHNESS5538 decision pages pre-date 2025TRUST6027 pages lack reviewer credentials
WHAT THIS MEANS FOR THE BUSINESSHelioForge is discoverable, but its most commercially valuable evidence is difficult to verify and quote.
Across the 24-question buyer set, the domain earns 19 of 72 possible provider citations. The largest recoverable gap is the “payback by country” cluster: only 2 of 18 responses cite HelioForge even though its benchmark contains the relevant figures.
PRIORITY 01 · 18 OBSERVATIONS
Country-level payback data is trapped in a PDF table
The benchmark page describes the methodology, but the country figures exist only inside a 42-page PDF. Providers repeatedly return IEA PVPS, EU Commission and SolarPower Europe pages with equivalent figures exposed in HTML.
Decision: publish one indexable country comparison hubPRIORITY 02 · 17 PAGES
Primary evidence is separated from 66 material claims
Claims name a percentage or payback period, but the publisher, study year and source link sit in a distant bibliography. A reader—or retrieval system—cannot reliably connect each claim to its provenance.
Five already-linked resources are blocked from retrieval
Seven research assets return 403. Five have external backlinks and two support high-intent query clusters, making access restoration more valuable than producing replacement articles.
Decision: send an exact URL remediation list to Web Ops
Recommendation order: unblock evidence → expose the country dataset in HTML → attach sources to claims → refresh dated incentive pages → rerun the same 72 observations. Do not commission a broad “solar trends” content programme until these assets are repaired.
OBSERVED OPPORTUNITY MAPFive query clusters explain all 72 provider observations.
The cited-response count is a simple observation: how often the target appeared in the cited sources returned for the fixed question set. It is not an estimate of total market share.
24 numeric claims; 18 lack an adjacent primary source. Country table exists only in the PDF.
Create an HTML table; show currency, unit, sample size, period and source on each row.Table extractable; 24/24 claims linked; canonical self-references.
HTTP 403 to the crawler; no HTML description or licence statement.
Remove the CDN bot block and add a landing page with field definitions and licence.HTTP 200; CSV downloadable; Dataset schema validates.
/authors/dr-mara-veenExpert-entity page42
Author name appears on 27 pages, but no biography, qualifications or reviewed-by linkage exists.
Publish credentials, subject remit, disclosures and links to reviewed research.Author entity resolves from all 27 pages.
WORKED IMPLEMENTATION BRIEF · PRIORITY 01
Commercial solar payback by European country
Decision: strengthen /research/commercial-solar-payback-europe. Do not create another general solar-payback article.
EXACT OUTLINE
Direct answer: median payback range and scope
Country comparison table with one row per market
Drivers: tariff, capex, incentives and self-consumption
Method, sample, exclusions and calculation formula
Limitations, review date and downloadable data
Standalone buyer follow-up questions
EVIDENCE SPECIFICATION
24 material figures; 24 adjacent sources
Currency, unit, country, period and sample on every row
Government or primary dataset for each incentive
Named analyst and independent reviewer
Method version and next review trigger
INTERNAL AUTHORITY PLAN
Link from /guides/lease-vs-buy-commercial-solar using “European solar payback data”
Link from /industries/warehouse-solar-roi using “country payback assumptions”
Link dataset and methodology pages back to the canonical hub
DEFINITION OF DONE
HTTP 200 and self-canonical
One extractable HTML comparison table
24/24 claims traceable beside the figure
Visible and schema dates agree
Same 18 questions rerun across the same providers
Later successful crawler request recorded
Each prescription is tied to the signal that triggered it, the query cluster it supports and a machine-checkable completion condition. Teams can assign the exported rows directly to editorial, research or web operations.
DIRECT SERVER-LOG EVIDENCE · 1–8 AUG 2026The logs isolate access failures that a page-only audit cannot see.
HelioForge supplied a seven-day edge-log export. IP addresses and raw lines were discarded before analysis. Every row below means “request declaring this user agent”—not independently verified network identity or proof of model training.
Five high-value research assets are discoverable but currently return 403.
ChatGPT-UserMEASURED312
28 unique pages · 2 latest failures
User-requested retrieval reaches guides but fails on two downloadable datasets.
Claude-SearchBotMEASURED908
61 unique pages · 7 historical errors, now resolved
Later 200 responses prove the earlier deployment failure is cleared.
Claude-UserMEASURED184
22 unique pages · 0 latest failures
Country and incentives pages have not appeared in this retrieval set.
PerplexityBotMEASURED1,105
69 unique pages · 0 latest failures
Strong guide coverage; original CSV and PDF assets remain unseen.
GooglebotMEASURED22,410
118 unique pages · 0 latest failures
General search discovery is broad, so the seven AI-crawler failures are not a site-wide outage.
VERIFIABLE NEXT MOVERemove the CDN rule blocking the seven research assets, then wait for later successful requests.
Definition of done: every affected URL records a later 2xx response from a recognized intended crawler. The action stays “ready to verify” until that evidence arrives through the next log import or secure event feed.
What this does not claim: the log evidence does not prove source-IP ownership, inclusion in training data, a citation, a recommendation, or downstream use. Those questions require different evidence classes.
LLM-BY-LLM COMPARISONObserved results stay separate from comparative estimates.
The first three providers were queried directly for this example. The remaining profiles show how the same measured evidence aligns with disclosed retrieval preferences; they are not presented as live citations.
MODEL FAMILYEVIDENCE CLASSRESULTWHAT HELD THE SITE BACKMOST RELEVANT CHANGE
OpenAIOBSERVED7 / 24 cited
Good retrieval of explanatory guides; weak selection of PDF-bound country data.
Expose country evidence in HTML with adjacent sources.
Google GeminiOBSERVED8 / 24 cited
Best current coverage; visible dates and government references influenced winning pages.
Refresh incentives content and preserve strong canonical signals.
Anthropic ClaudeOBSERVED4 / 24 cited
Returned research-led sources with clearer methodology and author credentials.
Connect every benchmark to method, author and primary evidence.
Microsoft CopilotMODELLED63 / 100
Indexable HTML and entity consistency are the largest comparative gaps.
Repair blocked assets and author entities first.
Meta AIMODELLED59 / 100
Short, self-contained answers are sparse on long commercial pages.
Add concise definitions and comparison summaries.
xAI GrokMODELLED58 / 100
Freshness and externally corroborated claims are the weakest matching signals.
Timestamp benchmark updates and cite current public datasets.
DeepSeekMODELLED56 / 100
Tables lack explicit units and explanatory field labels.
Publish semantic HTML tables with definitions.
Alibaba QwenMODELLED55 / 100
Regional comparisons lack consistent country and currency context.
Normalise geography, currency and period on every metric.
Moonshot KimiMODELLED54 / 100
Long reports are available, but executive claims do not link to supporting sections.
Add claim-to-section anchors and a structured contents map.
MistralMODELLED60 / 100
European relevance is strong; transparency and document access remain limiting.
Prioritise EU-source provenance and accessible report formats.
Observed total: OpenAI 7 + Gemini 8 + Claude 4 = 19 citations from 72 responses. Modelled scores are diagnostic comparisons only and are never added to the observed citation rate.
COMMERCIAL IMPACT PORTFOLIO · BASE SCENARIOWhich GEO work deserves the next 96 team hours?
Search demand comes from an illustrative 30-day Search Console export. Value per visit (€8.40) comes from illustrative GA4 revenue divided by sessions. Target CTR, 75% gross margin, €85 hourly cost and effort are visible planning assumptions—not measured outcomes.
CTR 3.74% · +290 visits/month€19,204 net · 706% ROIExpose the PDF-bound country dataset in one canonical HTML hub.
02 · Germany–UK incentivesCreate · 32h
8,100 impressions · 0.9% CTR · 1/12 cited
CTR 3.57% · +216 visits/month€13,606 net · 500% ROICreate the missing comparison owner from revalidated government evidence.
03 · Commercial PPA pricingStrengthen · 16h
6,800 impressions · 1.8% CTR · 4/15 cited
CTR 3.88% · +141 visits/month€9,296 net · 684% ROIAdd quarterly ranges, geography, contract term, caveats and source dates.
04 · Lease versus buyProtect + strengthen · 16h
9,200 impressions · 3.1% CTR · 7/15 cited
CTR 4.34% · +114 visits/month€7,258 net · 534% ROIPreserve the cited owner; expose financing assumptions and decision thresholds.
NEXT · Warehouse ROIStrengthen · 16h
5,200 impressions · 2.3% CTR · 5/12 cited
CTR 4.06% · +91 visits/month€5,516 net · 406% ROIStart after capacity frees; do not displace the four higher-value workstreams.
CONSERVATIVE35% of the CTR gap
Used for downside planning.
BASE65% of the CTR gap
Used for the investment queue above.
UPSIDE100% of the CTR gap
A ceiling scenario—not a target guarantee.
ANTI-INFLATION RULEOne demand row, one question
Overlapping query matches cannot be counted twice.
How the decision is defended: every row retains the allocated search queries, crawl evidence, provider checks, competitor rank, target page, confidence factors, implementation cost, formula and acceptance test. If GA4 value or explicit conversion assumptions are absent, IndexHalo shows visit scenarios but withholds revenue, ROI and payback.
CONTROLLED GEO EXPERIMENT · COUNTRY PAYBACK HUBPositive movement, with a directional confidence warning.
The primary target was met on the unchanged citation test. The sample is still too small for IndexHalo to call the provider movement statistically reliable, so the recommendation is to retain the release and collect two more cycles.
CALCULATED OUTCOMEPOSITIVE+27.8 percentage points
Target: at least +20 points
FROZEN HYPOTHESISIf HelioForge exposes its country-payback table in semantic HTML and places primary evidence beside each figure, observed citation coverage for the fixed country-payback question set will rise from 2/18 to at least 6/18.
Primary metric: timestamped provider citation coverage · Review window: 8 August–7 September 2026 · Owner: Growth lead
BASELINE · 8 AUG2 / 18 cited
11.1% coverage
Page readiness: 72/100
Search CTR: 1.4% from 12,400 impressions
Public competitor rank: 3 of 4
IMPLEMENTED CHANGEOne bounded release
HTML country table, adjacent primary sources, units, period, sample, methodology link and self-canonical.
The target threshold was exceeded, but 18 checks per period is below the 30-per-period reliability floor. This is a positive directional result, not a universal provider claim.
BUSINESS DECISIONKeep the new hub live; extend the fixed test for two more cycles.
Protect the target URL, add no competing country-payback page, monitor whether the citation gain persists, and validate the 1.8-point measured CTR increase against another equivalent window.
Causality boundary: the matching evidence shows an association between the bounded release and later movement. It does not prove that the release alone caused every change, reveal private ranking systems, or guarantee future citations. The experiment record and every evaluation remain exportable.
LONGITUDINAL CITATION OBSERVATORY · THREE COMPARABLE RUNSHelioForge gained visibility, but one commercially important provider loss still needs recovery.
Every movement below uses the same 24 buyer questions and the same OpenAI, Gemini and Claude provider set. Runs with changed denominators are excluded rather than blended into the trend.
IEA PVPS · EU CommissionProtect the HTML hub and extend the unchanged experiment for two cycles.Lease versus buy commercial solar7/15 → 8/15
OpenAI gained · Gemini gained · Claude lost one answer
Bank guidance replaced the target on ClaudePreserve overall gains; compare the bank’s financing assumptions with the target and recover Claude without changing the query.Germany versus UK incentives1/12 → 3/12
OpenAI cited · Gemini cited · Claude not cited
Government portals · legal summariesFinish the dated eligibility table and attach every rule to its primary programme source.
RETURNED-SOURCE LEDGERSource movement reveals who gained distribution.
Provider source sets remain fragmented. Consensus is strongest around IEA PVPS and EU primary sources, so HelioForge should preserve those adjacent references.
IMMEDIATE RECOVERYRestore Claude visibility for lease-versus-buy
Inspect the exact bank guidance that replaced HelioForge, expose comparable financing assumptions, then rerun the unchanged Claude checks. Do not rewrite the winning OpenAI or Gemini page.
EVIDENCE BOUNDARY2 runs excluded
One used a different question portfolio and one lacked verified provider evidence. Both remain stored but do not affect the movement calculation.
Scope: the 42/100 visibility index summarises this fixed monitored sample. It is not general AI market share, proof of training inclusion, or a guarantee that every user receives the same response. The full provider, question, URL and source ledger is downloadable as CSV.
BRAND NARRATIVE OBSERVATORY · HELIOFORGE RESEARCHThe brand is gaining citations faster than it is gaining clear attribution.
Exact aliases “HelioForge Research” and “HelioForge” were evaluated in the 72 stored provider answers. Naming, citation and recommendation context are reported separately so the team can see whether evidence distribution is strengthening the intended entity.
OBSERVED BRAND PRESENCE44/100+11 points since July
Protect the named benchmark owner; repair two uncited brand comparisons.Google Gemini14 / 2412 / 246 / 24
European data · current · authoritative
Keep dates and government provenance adjacent to the cited country table.Anthropic Claude8 / 247 / 242 / 24
Research-led · detailed · methodology concern
Expose the full method and version record in HTML; do not alter the winning data.
QUESTION × NARRATIVE DECISIONSThe report shows the words behind each action.
NAMED, NOT CITED · OPENAI + CLAUDEWhich commercial solar benchmark is most useful for European planning?
“HelioForge is a useful European benchmark, although its country methodology is difficult to verify from the public page.”
Evidence diagnosis: positive utility language, but the reason for selection is not attached to a citeable method page.
Action: add dataset version, sample, calculation method, limitations and direct primary-source links beside the country table. Repeat this exact question; target named + cited in both providers.
CITED, NOT NAMED · GEMINIWhat is the payback period for commercial solar in Germany?
“The 2026 European commercial payback table estimates a 5.8–7.1 year range.”
Evidence diagnosis: the correct HelioForge URL was cited, but the answer did not connect the figure to the publisher entity.
Action: place “Published by HelioForge Research” in the visible table caption, Dataset structured data, author record and download metadata. Verify both the URL and brand name appear.
NAMED + CITED + RECOMMENDED · OPENAIWhere can buyers compare commercial solar ownership models?
“HelioForge Research provides a practical lease-versus-buy framework with explicit financing assumptions.”
Evidence diagnosis: brand, page, useful attribute and differentiating evidence are aligned.
Action: protect this canonical URL and its financing table; monitor wording drift without rewriting a working owner.
FACTUAL REVIEW QUEUEThree statements require human verification before the next campaign.
These are flagged because they contain a date, number, absolute comparison or controlled risk term. IndexHalo does not label them false.
HIGH · VERSION CONFLICTClaude · country payback query
“HelioForge’s 2024 dataset covers 18 European markets.”
Owned evidence currently labels the downloadable benchmark “2026 edition · 21 markets,” while an older PDF remains indexable.
Owner: Research operations · retire or canonicalise the obsolete PDF; publish a version history; rerun the unchanged Claude question.MEDIUM · ABSOLUTE WORDINGOpenAI · PPA pricing query
“HelioForge offers the most comprehensive European commercial PPA benchmark.”
“Most comprehensive” is comparative and the response supplies no scope test.
Owner: Brand + legal · define a defensible coverage statement or remove the unsupported superlative from owned copy.MEDIUM · UNIT CHECKGemini · warehouse ROI query
“The benchmark reports savings of €142 per square metre annually.”
The canonical table shows €142 per installed kWp, making the returned unit important to verify.
Owner: Research editor · make the unit explicit in the heading, table schema and surrounding sentence; rerun the same question.
01 · ENTITY REPAIRUnify “HelioForge,” “HelioForge Research,” and the publisher organisation.
Add one stable organisation name, alternateName values, publisher relationships, logo, sameAs references and consistent PDF metadata. Check all seven cited-without-name answers after release.
02 · EVIDENCE ATTRIBUTIONConvert five named-but-uncited mentions into evidence-backed attribution.
Map each response excerpt to the exact owning page, add adjacent primary evidence and methodology, then test the same query/provider cells. Target: at least four of five become named and cited.
03 · NARRATIVE PROTECTIONPreserve the attributes already associated with winning evidence.
Do not manufacture praise. Keep “evidence-led,” “European data,” and “practical” defensible through current datasets, visible methods and decision-ready examples.
Evidence boundary: the 44/100 brand-presence index summarises exact alias, citation, recommendation-language and disclosed lexical signals in this fixed sample. It is not general market sentiment, private training evidence, a guarantee, or proof that a flagged statement is false. Every provider response, excerpt, rule and repeat test remains available in the Professional export.
BUYER QUESTION RESEARCH · JULY 2026The fixed portfolio covers only 64.9% of measured search demand.
IndexHalo combined the 24 monitored buyer questions with 162 additional demand phrases and audience questions found in the supplied Search Console export and 128-page crawl. Search impressions remain search evidence—not invented AI prompt volume.
PRIORITY OMISSIONS7not monitored
Each has ≥2,000 measured impressions, commercial or risk intent, and a weak or missing owner.
30% demand · 20% commercial intent · 20% ownership gap · 15% provider gap · 10% public competition · 5% trend
BUYER QUESTIONSCOREMEASURED DEMANDCURRENT OWNERPROVIDER EVIDENCEDECISION + TEST
commercial solar payback Germany warehouseCommercial · decision · GSC query
88/100CRITICAL
5,800 impressions · 46 clicks · 0.79% CTR · position 8.9 +42% vs June · Germany · desktop + mobile
Content gapNo page clears 20% relevance.Not observedOutside the 24-question provider set.Create one Germany warehouse payback owner with building assumptions, tariff period, financing case, units and primary sources. Add this exact phrase to monitoring; verify ≥60% owner relevance plus unchanged provider checks.
solar PPA price per MWh GermanyCommercial · decision · GSC query
82/100CRITICAL
4,300 impressions · 52 clicks · 1.21% CTR · position 7.4 +18% vs June · Germany · desktop
Ownership conflictPricing guide 41% · glossary 36%.Not observedReturned-source field unknown until activation.Consolidate the glossary definition into the pricing owner; publish quarterly €/MWh ranges, term, geography and exclusions. Activate the question and preserve the first provider baseline.
solar lease vs buy tax treatment UKComparison · consideration · GSC query
77/100CRITICAL
3,600 impressions · 41 clicks · 1.14% CTR · position 9.1 +9% vs June · United Kingdom
Weak ownerLease guide 38% relevance · 74/100 page.1/6 citedTwo providers return HMRC and bank guidance.Preserve the lease guide URL; add a dated ownership-and-tax comparison reviewed by a qualified UK specialist. Verify owner relevance, exact citations and equivalent search-period CTR separately.
2,900 impressions · 28 clicks · 0.97% CTR · position 10.6 +31% vs June · Germany + UK
Content gapThree unrelated pages each score below 18%.Not observedNo provider or brand narrative evidence.Create a risk checklist with jurisdiction boundaries, legal-review disclosure and contract clauses explained without legal advice. Activate only after the canonical owner passes evidence review.
2,400 impressions · 19 clicks · 0.79% CTR · position 11.2 NEW vs June · Spain · mobile
Weak ownerEurope hub 31% relevance · last reviewed 2024.0/3 citedGovernment and legal sources replace HelioForge.Publish a Spain section with programme status, eligibility, authority, dates and review trigger. Repeat the exact question across the same providers; do not count a failed response as uncited.
Question-level fuzzy matches remain inspectable, but each imported search row contributes to only one intent and portfolio summary.
ACTIVATIONPromotion creates a new baseline
Adding a question versions the fixed portfolio and requests a recrawl; prior runs remain stored but are excluded from unlike-for-like movement.
BOUNDARYNo invented prompt volume
61,300 is imported Google Search Console exposure for July—not model queries, total addressable demand, or a forecast.
Recommended decision: use the final available portfolio slot for “commercial solar payback Germany warehouse.” Put the other four candidates into the evidence backlog, then replace lower-value monitored questions only through a documented portfolio-version review. This preserves the 25-question limit and prevents low-confidence discoveries from diluting provider measurement.
ANSWER RETRIEVAL LAB · 24 FIXED QUESTIONSNine questions have a complete, self-contained answer unit.
IndexHalo inspected 1,486 bounded passages across 121 public pages. It preserved the exact text, page, section, evidence links and component scores for every question. This diagnoses retrieval readiness; it does not claim that a provider used the passage.
ANSWER-READY COVERAGE37.5%9 of 24 questions
6 unsupported · 5 fragmented · 4 without an answer unit
PASSAGES INSPECTED1,486Paragraph and sentence windowsANSWER-READY9Complete + evidenced + self-containedUNSUPPORTED6Relevant claim, no adjacent sourceFRAGMENTED5Depends on surrounding copyNO ANSWER UNIT4Below 25% relevanceINTERNAL CONFLICTS3Two pages within 8 points
QUESTIONEXACT TOP PASSAGEDIAGNOSISACTION + ACCEPTANCE TEST
UNSUPPORTED · 58/100commercial solar payback Germany warehouse5,800 July search impressions · fixed provider monitoring pending
“This result falls to 6.8 years when self-consumption exceeds 71% and annual electricity inflation remains above 3%.”Warehouse economics · /research/commercial-solar-payback
“This result” has no named subject. The assumptions are precise, but the paragraph has no adjacent source and cannot explain its baseline independently.
Rewrite the opening as “A 12,000 m² German warehouse reaches a modelled 6.8-year solar payback…”; state system size, tariff period, capex, financing and source beside it.Done when one passage is 60–180 words, relevance ≥55, provenance ≥45 and no competing page is within 8 points.
ANSWER READY · 82/100solar lease vs buy tax treatment UK3,600 July search impressions · 1/6 observed citations
“For a UK trading company, buying normally places the qualifying solar equipment in the capital-allowance review, while a lease usually treats periodic payments as operating expenditure; eligibility depends on contract substance and current HMRC rules.”Lease versus buy · /guides/uk-solar-lease-buy · HMRC source adjacent
The subject, jurisdiction, comparison and limitation travel with the excerpt. A named reviewer and dated HMRC source are adjacent.
Protect this canonical passage and update the tax review date without splitting the answer across new pages.Verify after every material HMRC change and repeat the unchanged six-provider sample separately.
NO ANSWER · 14/100industrial solar incentives Spain 20262,400 July search impressions · 0/3 observed citations
“European incentive programmes vary by market and should be reviewed before investment.”Europe overview · /guides/europe-incentives · last reviewed 2024
The only match is generic, two years out of scope, names no Spanish programme and cannot support an eligibility decision.
Publish a Spain 2026 owner with programme status, authority, eligible entity, project window, benefit, exclusions and next review trigger.Done when the new passage names Spain and 2026, cites the administering authority and clears 65/100.
CONFLICT · 63/100solar PPA price per MWh Germany4,300 July search impressions · pricing guide 67% / glossary 63% relevance
“German commercial solar PPAs observed in Q2 2026 ranged from €71 to €94/MWh…”/research/ppa-pricing-q2-2026
The glossary separately says “typical contracts range from €62 to €88/MWh” without a period. Both passages appear credible and answer the same question.
The score is withheld from answer-ready because two pages are within 4 relevance points and the ranges conflict.
Keep the dated research page as owner. Replace the glossary range with a definition and contextual link to the live benchmark.Done when the owner leads by ≥25 relevance points and all ranges share period, geography and methodology.
CONTENT SPECIFICATIONWhat the replacement passage must contain
Subject + direct conclusion + scope + denominator + units + timeframe + assumptions + adjacent primary source + visible review date + explicit limitation.
HISTORY CONTROLWhat changed since the previous crawl
Passage fingerprints detect a changed or lost top answer. One former answer-ready unit moved to “fragmented” after its opening sentence was removed.
SEPARATE OUTCOME TESTWhat is tested after publication
First recrawl the exact question and passage thresholds. Then, when server providers are available, repeat the unchanged provider set and store returned URLs independently.
Recommended decision: repair the German warehouse payback unit first. It combines the largest measured demand, high factual specificity and a bounded editorial fix. Do not publish another page; strengthen the existing owner, attach its primary model evidence, and verify the exact passage before interpreting any later provider movement.
EVIDENCE KNOWLEDGE GRAPH · 121 PUBLIC PAGESOnly 41.1% of material claims connect directly to their evidence.
IndexHalo created explicit relationships between 121 pages, 112 quantitative claims, 38 external source domains, 19 authors and 47 JSON-LD entities. Every edge comes from visible text, a published link or a structured declaration.
4,300 impressions“German commercial solar PPAs ranged from €71 to €94/MWh in Q2 2026.”/research/ppa-pricing-q2-2026 · Price range
UNSUPPORTEDNo external source link on page
Link the underlying transaction dataset or publish HelioForge’s collection method, sample, exclusions and version.
Verify: exact source URL retained; claim and Dataset publisher connect to one organisation @id.
2,400 impressions“Spain’s 2026 industrial solar incentive covers up to 35% of eligible project cost.”/guides/europe-incentives · Spain
UNSUPPORTEDPage reviewed in 2024
Do not refresh the date alone. Verify the programme with the administering authority, eligibility window and current legal review.
Verify: current authority source adjacent; visible and structured dates agree; no factual-review alert remains.
EVIDENCE LOST3 claims
A template migration removed adjacent footnote links while leaving the numbers visible. Restore the exact sources before the next campaign.
EVIDENCE GAINED11 claims
The new HTML benchmark connected country figures to primary datasets. Preserve these edges through future publishing changes.
GRAPH ACCEPTANCE TARGET≥80%
At least 90 of 112 material claims adjacent; zero unresolved value conflicts; one stable organisation identity across priority pages.
Recommended decision: treat the warehouse payback conflict and three evidence-loss regressions as release blockers. They touch the highest-demand buyer question and can propagate inconsistent answers. Assign Research Operations and Web Operations jointly; do not begin new content production until the canonical claim ledger and entity identity pass the stated recrawl tests.
GA4 SESSION-SOURCE ATTRIBUTION · 1–31 JULY 2026AI answer products delivered €38,720 in identifiable revenue.
IndexHalo found 1,284 sessions whose supplied source dimensions explicitly named an AI answer product, then joined each landing page to its crawl, passage and claim evidence. No direct, organic or unknown session was reclassified.
PERIOD MOVEMENT+68.6%€22,960 → €38,720
Equivalent 30-day windows
AI SESSIONS1,284+52.5% · from 842SHARE OF ALL SESSIONS3.80%1,284 / 33,789CONVERSIONS463.58% conversion rateREVENUE€38,720€30.16 per AI session
Commercial performance remains strong, but Claude lost one fixed citation. Preserve the answer, refresh HMRC evidence, then rerun the same provider test.
Retain page-level tracking and promote any URL above 100 sessions into the governed evidence register.
01 · GROWTH + EDITORIAL · 7 DAYSMove the benchmark PDF journey into a canonical HTML evidence hub.
172 identifiable AI sessions produced only 3 conversions and €12.27 per session, versus €35.66 per session on the payback hub. Preserve the source table and PDF download; add a decision-specific next step beside the answer.
Verify on the next equivalent GA4 source window: PDF/HTML combined sessions retained, conversion rate above 3%, revenue per session above €25, and no loss of answer readiness.02 · POLICY + GROWTH · 5 DAYSRepair the Spain answer before optimizing its conversion path.
118 AI sessions reached a page reviewed in 2024, but only one converted. Replace the generic Europe paragraph with a current Spain-specific answer sourced to the administering authority.
Verify: current source adjacent, passage ≥65/100, no stale-evidence alert, then test the qualified consultation CTA against the existing version.03 · ANALYTICS · NEXT IMPORTStandardize answer-product source names and keep dark traffic out.
Use landing page + session source / medium + engaged sessions + key events + revenue. Preserve ChatGPT, Perplexity, Gemini, Copilot and Claude rows; never distribute direct traffic across them.
Verify: current and previous windows have equal lengths, row totals reconcile to GA4, and provider plus landing-page sums both equal the AI totals.
Attribution boundary: this worked example is identifiable session-source attribution, not universal multi-touch attribution. A referred visit does not prove a preceding citation; an observed citation does not guarantee a visit or conversion; and revenue movement does not prove that GEO work alone caused it. The report keeps all three evidence sets visible and separate.
INTERNATIONAL GEO COMMAND CENTRE · SIX OBSERVED LANGUAGESGermany owns the strongest evidence. Spain has the largest commercial localization gap.
IndexHalo compared 121 public pages in English, German, Spanish, French, Italian and Dutch, retained 486 hreflang declarations, and joined each language inventory to measured search and identifiable AI-referral outcomes.
01 · SPAIN · POLICY + EDITORIALReplace the 2024 Europe summary with a Spain 2026 evidence owner.
The Spanish set trails German by 22 GEO points, receives 8,900 measured impressions and 164 identifiable AI sessions, but only 7 of 17 pages are source-ready.
Done when the Spanish alternate has current administering-authority evidence, score ≥70, reciprocal hreflang, and no claim copied from another jurisdiction without local validation.02 · WEB OPERATIONS · 3 DAYSRepair five broken alternate relationships.
Two French pages lack German return links, one Spanish target is not crawlable, and two regional codes point to pages whose analysed language disagrees.
Verify by recrawling the complete 28-set graph: every target returns 200, self-references, reciprocates, matches its language and remains self-canonical.03 · ANALYTICS · NEXT WINDOWCompare market outcomes without double-counting global pages.
Keep search rows and AI referrals attached to the measured landing-page language. Do not assign global English demand to every translated market.
Verify that language totals reconcile to page rows and that the next equivalent window preserves country, landing page and source dimensions.
International boundary: “not observed” does not mean IndexHalo recommends entering that market. Translation parity does not require identical copy. Claims, currency, law, dates, terminology and sources must be locally valid, and hreflang does not guarantee indexing or citation.
ANSWER SOURCE INTELLIGENCE · 72 VERIFIED CHECKSHelioForge owns 29 answer appearances—but primary authorities lead the highest-value questions.
The network separates owned pages, configured competitors, primary evidence and independent publishers. It shows what each provider returned, where the same domain persists, and which intervention is appropriate for each question.
OWNED SOURCE AUTHORITY74/100 in this monitored set
29 appearances · 24 questions · 3 providers
RETURNED DOMAINS38SOURCE APPEARANCES112CROSS-PROVIDER11QUERY GAPS14GAINING SOURCES5CONCENTRATION1,184 HHI
Finding: HelioForge states the result but withholds the sample and calculation notes that returned sources expose.
Action: publish the versioned methodology, sample, exclusions and worked calculation; give independent publishers the auditable dataset.
Verify: methodology passage ≥65/100 and the next fixed run records exact source movement separately by provider.
RECOVERY · PROVIDER LOSSLease versus buy for UK logistics propertyClaude lost HelioForge · GOV.UK gained
Finding: the owner remains technically accessible, but its finance assumption and UK tax source are stale.
Action: refresh the owned comparison and attach the current GOV.UK record. Preserve rather than replace the working canonical URL.
Verify: recrawl, confirm source adjacency, rerun Claude with the identical question, and record the returned URLs.
OWNED CONTENTStrengthen the answer asset when the owner exists.
Repair passage completeness, provenance, method and canonical stability. Do not create another overlapping page.
PRIMARY / DOCUMENTATIONPut authoritative evidence beside the exact claim.
Use official records, original studies and versioned documentation with dates, scope, units and limitations.
EARNED MEDIAMake original evidence independently reviewable.
Provide datasets and expert context where editorially relevant. Never buy, fabricate or imply an endorsement.
Source boundary: an appearance in a returned answer does not prove a domain is trustworthy, a backlink exists, the source influenced every user, or the source caused the answer. Authority is a disclosed score inside this fixed evidence set. Every URL must be reviewed before editorial, research or outreach action.
ANSWER CONSENSUS + CLAIM RISK · 72 VERIFIED ANSWERSThree value disagreements need research review before HelioForge changes its public evidence.
IndexHalo segmented 168 exact provider statements into 31 comparable claim clusters, retained every response-level source URL, and mapped each cluster to claims on the owning page. Agreement, disagreement and omission remain evidence labels—not declarations of truth or error.
ANSWER CONVERGENCE57%across the fixed provider set
OpenAIThe audited median payback period is 6.8 years across 84 sites.2 response-level source URLs · HelioForge cited
AnthropicThe audited median payback period is 7.4 years across 84 sites.3 response-level source URLs · HelioForge not cited
GeminiThe audited median payback period is 6.8 years across 84 sites.2 response-level source URLs · HelioForge cited
Decision: do not rewrite the headline from provider wording. Check the 84-site workbook: 6.8 years is the 2026 audited median; 7.4 is the 2024 baseline retained in a secondary PDF. Add explicit period labels and canonicalise the 2026 result.
Done when both values declare year, cohort, geography and method; the stale PDF links to the current record; repeat checks retain exact wording and URLs.
OpenAIEligible installations may receive support of up to 15%.Spanish authority returned
AnthropicThe programme can cover up to 20% of qualifying cost.Regional programme returned
GeminiRates vary by region, project size and applicant category.National and regional sources returned
Decision: both percentages can be valid. Replace the single-rate answer with a jurisdiction table separating national, regional and applicant-specific rules.
Done when each row names authority, region, eligibility, cap, effective date and source; legal review signs off before retesting.
OWNED CLAIMThe 2026 programme covers commercial projects between 50 kWp and 500 kWp, subject to regional eligibility.Adjacent administering-authority evidence on the public owner
OBSERVED ANSWERSNo provider states both capacity boundaries; two give generic eligibility language and one discusses residential installations.Omission is not proof of suppression or provider error
Decision: move the capacity range into the opening Spanish answer block, preserve the authority beside it, and remove the unrelated residential paragraph from the commercial owner.
Verify with the same Spanish question and provider set; compare exact statements, citation URLs and target presence separately.
PROTECT ACCURATE CONSENSUS
Benchmark scopeAll three providers correctly state that the study covers 84 completed warehouse projects.
Method boundaryAll three distinguish audited realised payback from modelled forward estimates.
Evidence formatTwo providers return the HTML evidence hub and the downloadable methodology workbook.
Interpretation boundary: agreement does not prove truth; disagreement does not prove error; omission does not prove suppression; and a response-level citation does not prove one URL supports every sentence. The research owner must inspect the exact source and scope before publishing a correction.
No. Core analysis is built into the platform, and customers never enter model-provider API keys.
Can it analyse and monitor a whole website?+
Yes. Professional accounts can crawl site-wide, schedule recrawls, compare snapshots, and track page and citation changes.
Can it prove that an LLM trained on a page?+
No independent tool can prove undisclosed training inclusion. IndexHalo reports public crawl evidence and timestamped citations without disguising estimates as private vendor facts.
Can subscribers cancel online?+
Yes. Signed-in subscribers can use Stripe’s customer portal for invoices, payment details, and cancellation.