Melissa Blake

Short version

Editor covering AI companion apps, character-chat platforms and the consumer AI tools around them since 2022, when the category first pulled away from general-purpose chatbots. Editorial lead for joi-ai.io. Byline on every content page across the site.

The beat

Consumer AI companion apps: Joi AI, CrushOn AI, Character AI, Replika, Kindroid, Ourdream, Lovescape, Chai, Talkie, PolyBuzz, Secrets AI. What each does well, what they charge, where the marketing diverges from daily-use reality, and which claims from a year ago are already out of date. The beat also covers adjacent NSFW-oriented platforms, adult AI content generation tools and how the payments, moderation and account systems around them work.

The pace of change in this category is faster than in general-purpose AI. A review from December 2025 saying Joi AI has no voice at all was accurate the day it was published and wrong six months later. Keeping a review usable in 2026 means going back and re-testing what shipped since. That's the work.

Background

Started writing about AI-driven consumer products in 2022, initially covering the shift from general chatbots (which had one voice, no memory, no persona) to character-chat platforms that could carry roleplay across sessions. The move from a service you asked questions to a service you had a relationship with was a change worth tracking, and the coverage evolved into the current focus on the AI companion niche.

Background before AI companion coverage was in consumer tech reviewing more broadly: subscription apps, mobile services, and the money-facing details (billing cycles, promo code behaviour, refund policy, hidden second-paywall patterns) that turn out to matter more than feature checklists. That grounding shows up in how Joi AI's Neuron layer gets covered on this site: the sticker price is the start, not the answer.

How I approach a review

Every review starts with primary sources: the platform's own pricing page, T&C, App Store listing and any pricing page the platform hosts. Anything the company controls about itself comes from the company's own pages first. What the company says about its price is what its price is; if a third-party review disagrees, the disagreement itself is worth flagging, but the platform's own page wins on what it charges today.

Independent review outlets that documented specific dollar figures in dated coverage are the second source. Names get cited when their reporting shapes a claim on this site. For Joi AI specifically, that means the mid-2026 reviews that documented Neuron pack prices, HELLO50 code behaviour, and live video call cost mechanics. Where those reviews landed, this site reports what they found, attributed.

User experience patterns come from Trustpilot and Reddit threads. These are treated as shapes rather than measurements. A repeated complaint about slow refund response times is fair to characterise; an isolated Trustpilot user claiming an exact number of days is not treated as generalisable data.

What I don't do

No fabricated personal measurements. If the site claims a specific latency figure, it comes from a named source or a live test done with documented method, not from my head. This rule cost some potentially attention-grabbing sentences on the homepage; keeping them out was worth it.

No star ratings or aggregate scores manufactured by this site. Trustpilot's 4.0 out of 5 is theirs; we cite it, we don't build a competing one. Readers make the call after reading the coverage rather than defaulting to a number I chose for them.

No accepting pre-publication review from the platforms being covered. Joi AI has never seen a draft of anything on this site before it went live. Independence from covered platforms spells this out.

Contact

Editorial contact: [email protected]. Corrections, source challenges, factual clarifications and press inquiries go there. Response time runs within a week for most requests, faster for material factual challenges that could change the shape of a claim on a page.

LinkedIn profile: linkedin.com/in/melissablake1. Bylines on this site: every content page, including the homepage Joi AI coverage and the supplementary pages listed below.

Location note

Based in Europe. Testing latency claims on the homepage reflect that: US-evening peak-hour observations come from published reviews rather than personal measurement, because a European IP hits the platform outside US peak. Where geography affects the claim, the site says so.

Bylines on this site

Joi AI coverage across the site sits under this byline. The main Joi AI coverage, about the site, review methodology, editorial policy, privacy policy, terms of service, age verification, and the 2257 statement all carry my editorial oversight. If you find something factually off on any of them, the email above is the right place to send it.