Products
Own how AI answers your products.
Product workspace
Products
Submitted products and analysis jobs for this tenant, loaded from persisted PostgreSQL reports and PIL drafts.
No products submitted yet
Add a product URL to generate the first AEO Report and PIL preview for this tenant.
Start first analysis| Product | Creator | Category | AEO score | Status | Updated |
|---|
Back to Products
Product evidence
Evidence workspace
Review sources and verify product claims without expanding the Products list.
Loading product evidence...
Single product analysis
Import Product URL
Run a URL analysis. Results are persisted to Products and can be re-run when content or scoring policy changes.
CSV batch import
Import multiple products
Upload a CSV with product_url and category. Optional columns: sku/external_id, and brand + model (fill these for marketplace URLs like Amazon /dp/ where the page blocks scraping — they replace the "Unknown Brand" fallback). Maximum 1,000 records per CSV.
Business unit settings
Settings
Manage account role, business profile, invitations, and the access model for this business unit.
Account
Current access
Loading account
Role and capabilities will appear here.
Business unit
Profile
Evidence scoring
Source weight policy
Weights flow through evidence contribution, TPTW, Trust Signal, Recommendation Readiness, and the weighted AEO score. Only matching evidence with confidence of at least 60% that has not expired participates; TPTW is capped at 100%. The final AEO score may remain at an active shortfall cap. Saving recalculates every current product report in this business unit.
Members
Invite by email
Access control
Business unit members
Permissions
Role model
Audit
Recent changes
Help center · complete product guide
How Hasmord Works
From the first product URL to evidence-backed AIO publishing, AI visibility testing, team access, and production monitoring.
Start here
From URL to measurable AI visibility
Hasmord keeps the source page, analysis, PIL, evidence, published AIO snapshot, and later measurements connected to one persisted product record.
- ImportSubmit one URL or a CSV catalog and choose the closest category.
- DiagnoseUse the Report and PDF to find answer gaps, weak trust, and unclear comparisons.
- StructureReview the PIL draft that turns page content into reusable product facts and buyer answers.
- VerifyAdd independent sources and run Fact Check before relying on important claims.
- PublishPublish a quality-gated AIO page with server-rendered content and JSON-LD.
- Measure and improveRun the Simulator, monitor Timeline and citations, then apply reviewed Agent changes.
ProductsSearch products...HY
24 Products87.4 Avg AEO91% PIL complete
Archer AX55router · completed
95.83ReportPILMore
Import and batch operations
Create reliable product records
An import is the start of a persisted workflow, not a disposable scan. A completed job creates the product, current report, PIL draft, and initial Timeline event.
Single URL
Enter an HTTPS product or service page, choose its category, and select Run analysis. Use Retry last run after a recoverable fetch failure.
CSV batch
Download the template and provide product_url and category. SKU/external ID, brand, and model are optional. One CSV supports up to 1,000 records and 1 MB.
Marketplace pages
Amazon and other marketplaces may block or vary rendered content. Supply brand and model in CSV, verify the extracted identity, and do not treat a blocked page as a completed product analysis.
Batch history
Filter imports by week, select one run, and review Completed and Failed rows separately. Each batch retains filename, importer, timestamps, SKU, row result, and job ID.
Retry and re-run
Retry failed rows processes only retryable failures. Re-run batch creates fresh analyses for the package and updates the corresponding Products records.
Batch exports
Download the overall PDF for remediation, or export failed rows as CSV for correction. The batch PDF summarizes every row; individual product PDFs remain available in Products.
Drop catalog.csv hereproduct_url · category · sku · brand · model
98 completed2 failed
SKU-AX55routercompleted
SKU-BLOCKEDmarketplaceretryable
Products, Report, and PDF
Turn a score into page changes
Product row
The row shows identity, SKU, creator when permitted, category, current AEO score, status, and updated time. Primary actions are Report, Edit, and Re-run; additional workspaces live under More.
Edit source
Correct the URL or category with Save & re-run. A plain save is intentionally unavailable because Report and PIL must match the corrected source.
Re-run analysis
Fetches the current page again and creates new report/PIL versions. The button shows a busy state; do not submit a second run while processing.
Delete product
Removes the product workflow after confirmation. Use deletion for genuinely unwanted records, not to refresh an analysis.
Priority fixes
Recommendations explain the missing buyer answer, why it affects AI recommendations, the exact page section to add, and estimated URS/AAR impact. They are guidance, not automatic proof.
PDF report
Downloads a branded remediation brief with the score breakdown, calculation, prioritized changes, and implementation guidance for content, product, or agency teams.
StructureMachine-readable page organization and product signals.
Query coverageDirect answers for category-specific buyer intents.
ExplainabilityClear reasons, conditions, constraints, and buyer fit.
ComparisonUseful alternatives and explicit tradeoffs.
Trust signalPolicy, support, first-party proof, and independent evidence.
FreshnessAvailability and time-sensitive facts that can be verified.
Final score calculation. Hasmord uses a weighted average plus shortfall caps. A strong total cannot hide a critical weak dimension; open Score calculation in the Report to see active caps.
82.7AEO
AEO REPORTEchoSubs AIWeighted average 88.4 · final score 82.7
100 Structure71 Coverage84 Explainability85 Comparison72 Trust
1
Add direct pricing and onboarding answersExplain cost, included limits, setup time, and first value.
Est. URS +2.8PIL, Evidence, and Fact Check
Separate product claims from proof
The Product Intelligence Layer (PIL) is the current structured product record. Evidence is stored separately so a claim can be traced, refreshed, weighted, and challenged.
PIL draft
Review product identity, category, commercial context, dynamic attributes, buyer fit, pros/cons, comparisons, FAQs, evidence references, trust/risk, and ranking signals. The JSON view is for integrations and advanced review.
Save PIL draft
Edits the persisted current PIL and recalculates dependent signals. Use it to correct facts or improve buyer answers; do not fabricate values missing from the source.
Evidence sources
Add merchant pages, first-party trust pages, manual verification, professional reviews, user reviews, or benchmarks. Record URL, supported claim, confidence, consensus, and optional validity date.
Source lifecycle
Active, expired, queued, processing, retry scheduled, completed, and failed states describe whether evidence currently counts and whether an automatic refresh succeeded.
Extracted facts
Open a source to inspect its normalized verified claims. Use Use for claim to create a Pro or Con bound to that exact evidence instead of editing raw JSON.
Fact Check
Compares saved PIL claims with active independent sources. Results are Consistent, Partial, Conflict, or Evidence needed; open details before changing a claim.
SourcesFact Check
BenchmarkConfidence 90%Consensus 82%
ev_techgearlab_ax55Independent throughput measurements at multiple distances.
Active4 Consistent1 Partial0 Conflicts3 Evidence needed
Confidence is not truth. Confidence describes source/extraction reliability; consensus describes agreement. Expired or low-confidence evidence can be excluded from TPTW even when it remains visible for audit.
AIO publishing and history
Publish a controlled, AI-readable snapshot
AIO quality gate
Checks buyer fit, limitations, tradeoffs, evidence, and at least three specific FAQs. Generic generated placeholders do not pass. Fix the PIL content, save it, then reopen AIO.
Publish / unpublish
Publishing freezes the latest reviewed PIL into a public server-rendered page with JSON-LD and a stable URL. Later draft changes are private until you publish again.
Public page
Includes product overview, key attributes, buyer fit, pros/cons, comparisons, evidence, FAQ, and freshness metadata. It is designed for crawlers and people, not as a replacement storefront.
Crawler observations
Records observed requests from GPTBot, PerplexityBot, ClaudeBot, and Googlebot. A request proves fetching, not indexing or inclusion in an AI answer.
Timeline
Shows immutable analysis, re-run, evidence/policy recalculation, PIL, publishing, Agent, and Simulator events with score/readiness changes and policy/version context.
Rank and AAR history
These fields remain “Not measured yet” until a successful Simulator run. Hasmord does not substitute score-derived proxy values for observed recommendation results.
AIO PAGEAI-readable product pagePublished
VisibilityPublicQuality gatePassedCrawlers / 30d3 of 4
https://app.hasmord.com/aio/t/.../products/...
GPTBotObserved · 4 requests / 30d
ClaudeBotNot observed
AI recommendation Simulator
Measure whether models recommend the product
Simulator is a controlled visibility test, not a live consumer search result or market-share estimate.
Buyer questions
Use suggested questions, add one realistic question per line, or leave the field blank for 15 generated category questions. Questions used in a completed run are saved with that run.
Three required models
Every full run sends the same 15 questions to Claude, Gemini, and GPT: 45 observations. A provider failure creates a partial audit result and does not update PIL ranking metrics.
Market context
The test can use up to 50 recently updated anonymous same-category Hasmord records plus each model's wider knowledge. Other tenants' identity, URLs, evidence text, ACL, and private buyer copy are removed.
Recommended in answers
The share of all model answers that included the target product. This is the observed Answer Appearance Rate (AAR) for this controlled run.
Average position / Top 3
Average position only counts answers where the product appears; lower is better. Top-3 rate measures first-three placement across all observations.
Model agreement
Shows how consistently the three providers treat the target across the question set. Open question details to see inclusion, rank, and target-focused reason by model.
AI recommendation visibility testcompleted
46.7% Recommended1.07 Avg position46.7% Top 376.7% Agreement45 Checks
Claude15 / 15 complete
Gemini15 / 15 complete
GPT15 / 15 complete
Safe optimization and monitoring
Review changes before they affect production
Agent Inbox
Run Agent Analyze converts current priority fixes into proposed change sets. Optionally re-enrich PIL first when source content has changed.
Safe Mode review
Open a proposal, inspect the target field and before/after content, edit the draft if needed, then approve and execute or reject it. Nothing is applied merely because the Agent suggested it.
Rollback
Executed change sets retain audit context and can be rolled back when a proposed change performs poorly or was approved in error.
Action attribution
Timeline connects approved changes with later AEO/readiness and Simulator observations so teams can distinguish correlation from an untracked manual edit.
Citation probes
Run controlled provider probes and review whether ChatGPT, Perplexity, Gemini, Claude, Copilot, or You.com cited the product. Probe results are synthetic checks, not analytics traffic.
Real AI referrers
When Google Analytics is connected and a property is mapped, Citations shows actual sessions sent by AI providers, top landing pages, and manual sync status. These are observed referrals, not inferred citations.
AGENT INBOX3 changes need reviewPricing answerPending reviewComparison blockApproved
CITATIONSReal AI referrers · 30d
ChatGPT38 sessions
Perplexity12 sessions
Gemini5 sessions
Business units, roles, and record access
Permission model
Role permissions define the default scope. Product permissions add explicit read, edit, or manage access to an individual record.
| Role | Best for | Product access | Administrative scope |
|---|---|---|---|
| Owner | Business owner or final administrator | All products, creators, reports, PIL, evidence, PDFs, and record permissions. | Full control; profile, members, API keys, evidence policy, owner transfer, and removal with product reassignment. |
| Admin | Operations lead or team administrator | All products and creators; can manage record access. | Invites and manages non-owner roles/API keys; cannot transfer owner or edit owner-only profile controls. |
| Manager | Product or content lead | All products and creators; can re-run and share records. | No member, API key, owner-transfer, or owner-only profile management. |
| Editor | Content, SEO, or catalog operator | Own products plus explicitly shared records. | Can import and edit/re-run records they own or can edit; no tenant administration. |
| Analyst | Reviewer, consultant, or analyst | Own products plus explicitly shared records. | Can submit analysis work and inspect permitted results; cannot edit/delete/share tenant-wide records. |
| Viewer | Read-only stakeholder | Only explicitly shared readable records. | Cannot import, re-run, edit, delete, share, or manage settings. |
Multiple business units
One identity can belong to several tenants. Use the account menu to switch; products, members, evidence policy, audit logs, and permissions always remain tenant-scoped.
Invite by email
Owner/admin selects an email and role. The recipient must accept using the invited email. Re-sending a pending invitation updates delivery instead of creating a duplicate.
Product permissions
Grant read, edit, or manage access per product. Owner/admin/manager retain tenant-wide product visibility even when a record ACL is changed.
Member removal
Before removing a member, select a new assignee for products they created or managed. This prevents orphaned records and preserves owner/admin access.
Owner transfer
Only the current owner can transfer ownership to an active member. Transfer before removing or demoting the current owner.
Evidence policy
Source-type weights affect TPTW, Trust Signal, recommendation readiness, and AEO score. Saving recomputes every current product and writes an audit event.
BUSINESS UNITCNNEX Product TeamOwner · [email protected]
HY[email protected]Analyst · 4 shared productsanalyst
MJ[email protected]Editor · 12 owned productseditor
Scenario playbooks
What to do in common situations
Open a scenario for the shortest production-safe path and its acceptance check.
01Establish a baseline for the first product
- Import the canonical product or service URL and select the closest category.
- Confirm identity, URL, category, and completed status in Products.
- Open Report; record the overall score, active caps, and first priority fix.
- Open PIL; correct only verifiable facts, then save.
- Download the PDF as the implementation brief.
Done when: the product survives logout/login, Report and PIL match the source, and the PDF opens with all sections.
02Improve low Query Coverage
- Open the first “Expand buyer-intent coverage” recommendation.
- Add crawlable sections or FAQs for the named missing intents, using direct answers rather than labels alone.
- Publish the source page and use Re-run.
- Compare the new Timeline event and score; do not edit PIL merely to inflate the score.
Done when: the page itself contains the answers and the new report detects them.
03Import and maintain a catalog batch
- Download the CSV template; keep each file at 1,000 rows or fewer.
- Include stable SKU values and brand/model for marketplace URLs.
- Upload, filter the batch by week, and open its detail.
- Correct exported failure rows, retry only retryable rows, then download the batch PDF.
- After a site-wide change, use Re-run batch and confirm Products reflects the new jobs.
Done when: completed + failed equals total and every expected SKU appears in Products or the error CSV.
04Handle a blocked or inconsistent marketplace page
- Verify the URL resolves to a real product and not a “Page Not Found” response.
- Provide brand/model in CSV or correct identity in PIL after verifying the listing.
- Add independent evidence for claims the marketplace page does not reliably expose.
- Retry only after the source is accessible; do not accept placeholder brand, price, or model values.
Done when: identity is correct and unsupported claims are either evidenced or removed.
05Add a benchmark and verify a claim
- Open Evidence → Sources and add the HTTPS benchmark URL, claim note, confidence, consensus, and validity date.
- Wait for extraction/refresh to complete, then inspect Extracted facts.
- Select Use for claim, choose Pro or Con, and save the evidence-backed claim.
- Open Fact Check and inspect Consistent, Partial, Conflict, or Evidence needed details.
Done when: the PIL claim references the evidence ID and Fact Check explains the relationship without raw source noise.
06Publish or update an AIO page
- Complete specific buyer fit, limitations, tradeoffs, evidence, and at least three useful FAQs in PIL.
- Save PIL and reopen AIO to refresh the quality gate.
- Publish, open the public URL in a signed-out window, and verify structured sections and favicon.
- After later PIL edits, publish again; the existing public snapshot does not change automatically.
Done when: visibility is Public, the public page loads without authentication, and its content matches the reviewed snapshot.
07Measure AI recommendation visibility
- Open Simulator and enter realistic category questions or use the automatic set.
- Run the 45-check test and wait for all three providers.
- Review inclusion, average position, Top-3, agreement, saved questions, and per-model reasons.
- Use Timeline to compare later runs after controlled content or evidence changes.
Done when: the run is Completed, has 45 observations, and rank/AAR appear in Timeline. Partial runs must not update PIL.
08Apply an Agent recommendation safely
- Run Agent Analyze from Agent Inbox.
- Open each change set; verify the target, before value, proposed value, and supporting fix.
- Edit if necessary, approve and execute only the selected change, or reject it.
- Re-run analysis and Simulator before attributing an improvement; rollback if the change is wrong.
Done when: the audit trail shows the decision and Timeline shows the resulting version.
09Separate probes, crawler fetches, and real AI traffic
- Use AIO crawler observations to confirm bots fetched the public page.
- Use Citations probes to run controlled provider citation checks.
- When GA is connected and mapped, sync and review actual AI-referral sessions and landing pages.
- Never describe any one of these as proof of the other two.
Done when: the report labels observed fetches, synthetic probes, and analytics referrals independently.
10Work with an analyst or agency
- Invite the person as Analyst for review work, or Editor if they must change owned/shared records.
- Share only the required products through Permissions.
- Confirm they selected the correct tenant and can see only expected records/actions.
- Revoke record access or remove the member with product reassignment when the engagement ends.
Done when: a cross-role test confirms least-privilege access.
11Change evidence weights
- Record the current policy version and representative product scores.
- Change only the intended source weights in Settings.
- Select Save & recompute and wait for the product count confirmation.
- Review Timeline and Report calculation on representative products; revert deliberately if results are not defensible.
Done when: the audit log, policy hash, and recalculated products agree.
12Remove a member without losing products
- Select an active replacement member in Settings.
- Reassign and remove the departing member.
- Confirm owner/admin still sees all products and the replacement can manage assigned records.
- Review the audit entry and revoke any API keys associated with the old workflow.
Done when: no record is orphaned and the removed account cannot access the tenant.
Statuses, messages, and recovery
Interpret the system before retrying
queuedAccepted and waiting for a worker. Leave the record in place.
processingFetch, extraction, scoring, or provider calls are active.
completedThe operation finished and persisted its current result.
partialSome providers or rows succeeded. Review details; do not treat it as a full measurement.
failedThe operation stopped. Read the specific error before retrying.
retry scheduledAn evidence refresh will run again at the shown time.
expiredThe source remains auditable but is excluded from current evidence weighting.
dead letterAutomatic retries are exhausted; correct the row/source and submit deliberately.
draft / privateThe AIO snapshot is not publicly readable.
published / publicThe reviewed snapshot is available at its public URL.
not observedNo matching crawler/referrer event was recorded in the stated window.
not measured yetThe metric requires a successful Simulator observation; no proxy is shown.
Authentication looks correct but an action says auth required
Refresh once, confirm the account panel shows the expected tenant, then sign out/in. If the session is valid but the action is absent, check role and record permission rather than repeating OAuth.
A re-run shows no score change
Confirm the published source HTML actually changed, the canonical URL is correct, and the new content is crawlable. Compare Timeline versions and extracted page facts before changing scoring policy.
AIO says FAQs are generic
Three cards are not sufficient by themselves. Each question and answer must be product-specific, direct, and useful; save PIL before reopening the AIO gate.
Evidence refresh returns 4xx/422
Open the source manually, confirm it is public HTTPS and not bot-blocked, correct the URL if needed, then retry. A scheduled retry does not guarantee the source will become fetchable.
Simulator fails or is partial
Do not infer rank. Check provider statuses and retry later; only a three-provider completed run writes ranking signals and Timeline observations.
A product is missing
Check the selected tenant, search filters, pagination, creator scope, and record permissions. Products are never automatically copied between business units.