Articos ✓

What's New

Since the v2 platform rebuild, Articos has added two new research types
(Message Testing and Landing Page Testing), a reusable Persona Library, and
dozens of workflow refinements across all five steps.

New Research Type: Message Testing

• Test up to 5 message, headline, or value-proposition variants against your
  personas and see which one resonates -- not just single-variant interviews
• Guided setup with message-type categories (headline, value prop, email
  subject, ad copy, and more) plus a couple of quick context questions
• Every persona is interviewed against every variant, so results are a true
  per-persona and per-segment comparison rather than a split sample
• Dedicated Step 4 view: persona list on the left, that persona's
  variant-by-variant responses on the right
• Report ranks a winning variant with a per-dimension scorecard; single-variant
  tests get a diagnostic "how does this land" read instead of forced
  winner/loser language
• Add another variant straight from a finished report (up to 5) and get a
  refreshed head-to-head -- it reuses the same personas and questions, so
  there's no starting over

New Research Type: Landing Page Testing

• Upload or capture a single landing page, pick one goal, and get persona
  feedback scored against that goal
• Single-page scorecard report -- goal-achievement score, UX-dimension
  breakdown, themes, and recommendations (not a two-way A/B comparison)
• Available on the free Trial, not just paid plans
• Interview-script edits in Step 3 now reliably persist through execution

Persona Library

• Save any persona -- or a whole audience at once -- to a personal library and
  reuse it across future studies instead of regenerating every time
• New Persona Library page with search, filter, and a full read-only detail
  view (bio, personality traits, attributes)
• "Reuse Existing Personas" option when starting a new study (idea validation,
  A/B testing, and message testing)
• Saved personas show a persistent "Saved" badge so you always know what's
  already in your library

Persona Generation & Step 2

• Refine a persona in plain English -- type an instruction ("make her more
  skeptical of new tools") and the AI regenerates just that persona; review a
  before/after diff before applying
• Persona names are now guaranteed unique and consistent with each persona's
  origin; gender is a proper generated attribute for better distribution
• Step 2 redesigned into a two-pane layout -- stats and per-role breakdown on
  the left, a filterable persona grid with a slide-in detail drawer on the right
• Add a single persona to a role without regenerating the whole set

Step 1: Intake & Onboarding

• Attach a document to your Step 1 chat (PDF, Word, TXT, or Markdown, up to
  10MB) and the AI reasons over it alongside your messages
• Upload your own interview script in Step 3 -- the AI extracts your questions
  and you choose to append them or replace the generated script
• One-click starter-example cards on the Step 1 opener for first-time users
• Revamped first-run onboarding: choose your role and primary use case right
  after signup
• Self-service "Forgot password" flow with a branded reset email and reset page

Reports (Step 5)

• Reports now generate in the background -- close the tab or refresh and the
  report keeps building server-side, ready when you return
• Live progress on the Step 4 card ("Generating your report" -> "View Report")
• Download your report as a PDF from Step 5
• Shared/public A/B report links now include the same participant count,
  question count, and variant screenshots as your private view

Interviews (Step 4)

• Download every completed transcript at once as a single combined PDF --
  works for standard, A/B, and message-testing studies
• Download an individual persona's transcript as a PDF from the interview modal
• Interviews are more resilient to transient AI hiccups: a mid-interview
  failure now retries automatically instead of failing outright, and a
  permanently failed interview shows a clear "failed" state instead of spinning

Plans & Billing

• New "Pay per research" option -- buy a one-time research pack (credits that
  never expire) instead of a monthly plan, right from the upgrade modal
• Trial extended to 7 days (up from 3)
• Study limits now count completed studies, not started ones -- create and
  abandon as many drafts as you like; only a finished study with a report
  counts against your plan or pack
• Trial users now see the full product UI, with upgrade prompts appearing at
  the point of use instead of grayed-out feature

All notable changes to Articos are documented here

v2 is a major platform rebuild introducing A/B testing research, and AI-powered persona interviews via Cloud Tasks, a structured report pipeline with charts and citations, deep persona generation, and a fully redesigned UI. The backend was migrated to Python across all primary Cloud Functions.

New Research Type: A/B Testing

• New ab-test research type for comparing two variants (e.g., landing page A vs. B)
• Variant preview section with live browsing animation during interview setup
• A/B comparison sections in the report with side-by-side analysis
• Screenshots of each variant proxied through Firebase Storage
• Goal selection limited to a single goal per study
• Auto-naming support for A/B test research

AI Persona Interviews (Step 4)

• Persona interviews now triggered via Cloud Tasks (fully async)
• Full interview status lifecycle: not-started > setting-up > in-progress > completed
• Interview questions chained via previous_response_id for coherent multi-turn conversations
• Research context and roleplay prompt injected into follow-up answers
• Medium reasoning effort enabled for interview AI
• View Transcript and Follow Up Question buttons on completed interview cards
• Preparing the state shown on the interview card during the setup phase
• Live browsing animation while AI interviews are running

Structured Report Pipeline (Step 5)

• Full structured report UI with charts, citations, and evidence drawers
• Report sections written in parallel for faster generation
• Pipeline resumes from the last completed stage on re-entry (no restart)
• Progressive loading states with skeleton UI and auto-scroll
• TOC sidebar with section navigation and ellipsis truncation on long titles
• Expert critique sections with correct prompts per research type
• A/B variant screenshots labelled as "Variant A" / "Variant B" in the report
• Report download and share buttons with PDF export hook
• Public report endpoint updated to serve structured pipeline reports
• Section IDs sanitised for Firestore paths
• Comparison matrix columns aligned when LLM includes label headers

Deep Persona Generation (Step 2)

• Progressive persona loading -- each profile updates independently as it completes
• Geocoding and dynamic persona attributes (country flags, world map pins)
• Archetype field rendered in persona card header
• OCEAN personality scores normalised to a 1-100 scale-
• Location and origin country rules added to the deep persona prompt
• Global location diversity enforced (no US default)
• Radar chart with secondary attribute dimensions from generate_relevancy_dimensions
• Persona card bio tooltip showing full text
• Edit persona modal hides AI refinement section when step is completed
• Edit/delete buttons disabled on persona cards after step 2 is completed
• Generation state persisted across step navigation

Auto Study Type Detection (Step 1)

• Research type (idea-validation, landing-page-feedback, ab-test) now auto-detected from conversation -- no upfront selector
• inferredStage returned from backend and consumed by frontend
• Conversation saving shifted from backend to frontend
• Profile selection is disabled after step completion to prevent race conditions
• Research names no longer include _study{n} suffix
• Row-based roles UI replaces card grid on ProfileCard

UI / Design System

• Full platform UI revamp (Step 1, Step 2, Step 3 redesigns)
• Design tokens replace hardcoded colors and arbitrary spacing values
• ActionTooltip component added and applied across the platform
• Follow-up chat panel restyled to Articos warm gold/cream design system
• Follow-up sidebar collapses on open
• Accessibility: aria-label added to all icon-only buttons
• Scatter chart uses dynamic domain; persona bio filter fixed

[v0.2.8] and earlier

Legacy releases prior to the v2 platform rebuild. See git history for details.