Marc Allard, AI companion reviewer

Marc Allard
AI companion reviewer
Active since 2023

Who I am and why this site exists

I'm Marc Allard. Since 2023 I've spent more time than any reasonable person should signing up for, paying for, and testing AI girlfriend apps, NSFW AI image generators, and AI roleplay platforms. Candy AI, Promptchan, SoulGen, Replika, Character.AI, Janitor AI, CrushOn AI, Muah AI, Joyland AI, Nomi AI, SpicyChat, Kupid AI, Eva AI, OurDream AI, Botify, Nastia AI, Seduced AI, and roughly fifteen others I would rather not name in polite company.

Most reviews of these products online fall into two categories. The first is thinly veiled affiliate fluff written by someone who clearly never logged in. The second is a one-paragraph summary that doesn't mention pricing, content policy, or any actual user experience. I started AI Companion Guide because I wanted a real review the way I wished one had existed when I was deciding which platform to subscribe to.

I work alone. I write everything on this site myself. When I cite Reddit threads or Trustpilot ratings I'm pulling them from the live source, not paraphrasing what some other review wrote. When I quote pricing I have a screenshot of my own credit card statement to back it up.


How I test (the actual methodology)

Every platform reviewed on this site goes through the same five-step process before publication. The bar is not "I poked around for an hour" — these are paid subscriptions used long enough to see what breaks.

  1. 14 to 30 days of active use on a paid account. Free tiers get tested separately. I don't accept comp accounts from vendors because the experience differs.
  2. Full feature exercise. Chat. Image generation. Voice. Video where it exists. Content policy boundaries. Cancellation flow tested at the end so I know how hard they make it to leave.
  3. Real spending tracked. I record what I spend each month including tokens, credits, and any "addons". Pricing claims on this site reflect what I actually paid, not the marketing page floor.
  4. Reddit and Trustpilot triangulation. My experience is one data point. I read 30-50 community threads and the Trustpilot complaint summaries to check whether what I saw matches what other paying users saw.
  5. Sit on the review for 48 hours before publishing. Some of my first-draft verdicts have softened, and some have hardened. The cooling-off pass catches both.

This is what separates a real review from an SEO page. If I get something wrong, I want to hear about it — see the contact section below.


What I test for (categories covered)

The site currently covers 20 platforms across four overlapping categories. Each one is reviewed as its own page; comparisons live within the relevant reviews.

Many platforms straddle categories (Candy AI does chat + image + video; Promptchan does image + chat + video). The category split is for navigation, not strict taxonomy.


Editorial independence

I use affiliate links when they exist for products I recommend. When you click an affiliate link and sign up, I may earn a commission at no cost to you. My verdicts are independent of whether a product has an affiliate program. Products with no affiliate program get recommended too when they're the right answer.

Some products on this site (Stable Diffusion, free-tier alternatives, open-source models) pay me nothing and never will. They're listed because they fit the use case.

What I will not do: write a review I don't believe to make a commission. If a platform is bad and pays a 30% affiliate cut, I will still tell you it's bad. The longest-term play for this site is being credible enough that people come back — pumping a bad product for one commission breaks that.

Full disclosure on the disclosure page.


Credentials and background

I'm not an academic. I don't have a Ph.D. in machine learning. What I have is three years of paying real money to test these products and the patience to keep doing it as the category changes monthly. New models drop, pricing shifts, content policies tighten or loosen. Most reviews you'll find online stopped being accurate six months after publication. I update mine.

Devices used for testing: MacBook Pro M2, iPhone 15 Pro, Pixel 8 Pro. I test mobile browser experience on iOS and Android separately. Cancellation flows tested from each device because some platforms make it harder on mobile.

Reddit accounts used for community triangulation: (pseudonymous, not shared publicly for obvious reasons).


Find me elsewhere

Cross-references for verification:


Get in touch

Corrections, factual disputes, missing products to test, or general feedback: email hello@aigirlfriendtested.com. I respond to every message within 7 days. If something I published is wrong, I want to fix it.

I do not accept paid placements, sponsored sections, "guest posts", or product reviews in exchange for free accounts. PR pitches asking for any of these will be ignored.