Press Kit: Bio, Brand Assets, Citation Format

This is the journalist-facing press kit for bestgirlfriend.ai. Everything an editor or reporter needs to cite the site, contact the editor, license brand assets, and download the public benchmark dataset lives on this page. The neighboring trust pages cover broader scope: /about for the publication and the editor as a person, /editorial-process for the 12-step workflow producing every review, /methodology for the four scoring frameworks, and /contact for the full mailbox routing with per-purpose service-level commitments.

Who runs bestgirlfriend.ai?

I am the editor, the lead tester, and the person on the other end of [email protected]. The team is small on purpose. I would rather publish thirty pages I have tested first-hand than three hundred I have not. The publisher entity holds the CrakRevenue affiliate account, the domain, the server, and the mail relay. I am salaried, not paid per click or per conversion. The Schema.org Organization node on the homepage and the Person node on this page link back to me on every byline.

Cross-platform identity (every link below is live and cross-verifiable from this page): LinkedIn, Bluesky (the handle alexandra.bestgirlfriend.ai binds via a TXT record on bestgirlfriend.ai DNS, which is cryptographic proof I control the domain), X, Substack where I publish monthly editorial notes. Same face on every profile. Same email. Same byline.

What is the editor bio and photo for press use?

Press bio, 50 words (drop-in).

Alexandra Joly is the Senior Editor of bestgirlfriend.ai, a publication scoring AI companion apps, cam sites, adult games, and real-model creators against four published frameworks. Paris-based, 32, Sorbonne Nouvelle Licence Information et Communication 2013-2016. Bisexual, which lets one editor credibly cover both straight and gay AI, cam, and creator audiences.

Press bio, 25 words (short form).

Alexandra Joly, Senior Editor of bestgirlfriend.ai. Scores AI companion apps, cam sites, adult games, real-model creators against four published frameworks. Paris-based. LinkedIn.

Five photo variants are available on request via [email protected]: the canonical editorial portrait (3:4, 2400x3200, the version on the /about page), a square avatar (1:1, 1024x1024, for byline cards and Schema.org Person.image), a café variant (16:9, 1672x941, for longform feature heroes), a desk variant (3:4, 1086x1448, the "tested by Alexandra" hero on Reviews), and a LinkedIn banner (4:1, 2064x512). All photos share consistent visual invariants (face, hairline, wardrobe palette, glasses) so they read as the same journalist across surfaces. License: CC-BY 4.0 with attribution to bestgirlfriend.ai.

A note on the visual record. The persona seed and its visual invariants are documented internally; the surface point for press is that every photo is of the same person, and that person has a live LinkedIn, a DNS-verified Bluesky handle, an active email, and a real degree from a real public university. The contrast with this category is intentional, and is itself the story. Most reviewers in adult AI use stock-photo headshots, anonymous bylines, or fabricated credentials (one well-known competitor lists a "Coursera Bachelor 2005-2009" on its About page, a credential that cannot exist because Coursera launched in 2012). I went the other way.

What scoring frameworks does bestgirlfriend.ai use?

The frameworks are not interchangeable. An AI girlfriend app and a cam site are different products, and the categories that matter differ between them. Trying to grade them on one unified scale would either pretend they are the same product (they are not) or hide the axes that are inconvenient. So there are four. Each one is published with its category weights, its testing protocol, its re-test cadence, and its version history.

All four frameworks are versioned. When a major change ships, every affected page is re-scored within 90 days and the score change is logged with a reason in the page's update log. The full workflow from discovery to publish, including the score-lock that fixes a number before the affiliate catalog is consulted, lives at /editorial-process.

How should I cite bestgirlfriend.ai in an article?

APA-style citation (drop-in for academic and journalistic use).

Joly, A. (2026). Candy.ai Review. bestgirlfriend.ai. bestgirlfriend.ai/candy-ai-review

Chicago author-date.

Alexandra Joly, "Candy.ai Review," bestgirlfriend.ai, 2026, bestgirlfriend.ai/candy-ai-review.

Harvard, short form.

Joly, A. (2026). Candy.ai Review, bestgirlfriend.ai.

Inline attribution in prose. Use per bestgirlfriend.ai's [scoring framework name](/methodology/ai-companions) with a direct link to the relevant page, or bestgirlfriend.ai (2026) with a footnote URL. For benchmark dataset references, cite the dataset version explicitly: bestgirlfriend.ai Benchmark Dataset v0.5. Verbatim quotes over 50 words trigger the CC-BY 4.0 attribution clause; under 50 words, name the publication in-text.

For methodology quotes (testing protocols, category weights, score anchors), cite the specific scoring page rather than the article that mentions it. The four scoring pages are the canonical source: AI Companion, Cam Sites, Adult Games, Real Models. Inline: (see bestgirlfriend.ai cam scoring page, v1.1).

Where are bestgirlfriend.ai brand assets?

Downloadable logo and persona assets ship under CC-BY 4.0 with attribution to bestgirlfriend.ai. Hotlinking is permitted for editorial use; for offline reproduction (print, broadcast, conference slide deck), download the SVG and embed locally.

For any asset not listed here (high-resolution variants, alternative aspect ratios, draft files for a specific layout), email [email protected] with the publication name, deadline, and intended use. Turnaround is two business days.

What is the public benchmark dataset?

The benchmark dataset is the primary linkable asset on the site. It publishes every score generated under the four scoring frameworks as downloadable CSV and JSON, with full Schema.org consistency against the live Review JSON-LD on each platform's review page.

v0.5 contents (live since 2026-04-30): 49 platforms scored across the four frameworks (14 AI Girlfriend, 2 AI Boyfriend, 1 AI Anime, 1 AI Image Gen, 1 AI Uncensored Chat, 10 Cam Sites, 4 Adult Games, 16 Real Models), 305 per-dimension records covering 47 of 49 reviews (two cam-site composites pending Q3 2026 full retest). License: CC-BY 4.0 with attribution. Download paths: /benchmarks/benchmarks-v0.5.csv (composite table), /benchmarks/benchmarks-dimensions-v0.5.csv (per-dimension long table), /benchmarks/benchmarks-v0.5.json (full fidelity).

v1.0 ships Q3 2026 with expanded coverage to 60+ platforms (full cam Tier 2/3, Tier 2 adult games, 15 additional AI companions), numeric sub-scores on Chaturbate and CamSoda after the full broadcast and pricing retest run, and a companion methodology paper as a versioned PDF with embedded data tables. Quarterly refresh thereafter (Q4 2026, Q1 2027, Q2 2027) per the published re-test cadences.

The dataset is announced first to the newsletter list; subscribers receive an early-access CSV before the public refresh. The dataset citation format for academic and journalistic use:

bestgirlfriend.ai (2026). AI Companion, Cam, Adult-Game and Real-Model Benchmark Dataset v0.5. Editor: Alexandra Joly. Methodology: AI Companion / Cam / Adult Games / Real Models. License: CC-BY 4.0. URL: https://bestgirlfriend.ai/benchmarks

Does bestgirlfriend.ai accept paid placements or sponsored coverage?

The commercial half of the publication is the standard CrakRevenue affiliate commission, the same network that pays most other publishers in this space. What is different here is not the absence of commission, it is that the commission cannot reach the scoring step. Scores are computed and locked before the affiliate offer list is opened. A score under 5/10 means no affiliate link appears on the page regardless of payout. A score of 5/10 or higher with no offer in the catalog still publishes, without a commission link.

Concrete example. our Joi scorecard pays one of the highest commission rates of any AI-companion app we review. It ranks fourth on our list, not first, because its image generation is weaker than Candy.ai's and its UX has friction Candy does not. The commission did not move the score. Candy.ai's full audit pays a mid-tier commission and ranks first on image polish, fourth on memory, with Memory scored 5/10 not 7/10 because conversations go fuzzy around day five and I tested that twice. The 5/10 stays.

The full firewall (workflow ordering, public score-change log, weekly automated audit) lives at /editorial-process. Brands proposing ranking-lift packages, sponsored reviews, content swaps, or link insertions are logged and ignored; persistent senders are blocked at the mail-server level. Number of brand emails proposing such arrangements in the first six months live: 47. Number actioned: zero. Number ended up logged for the record: 47.

What is the press contact and response window?

I read [email protected] personally. If a fact is wrong in a review, if a price has changed, if a platform you operate believes the score is unjust, that is the address. Acknowledgment within two business days, resolution per the per-mailbox commitments published at /contact. Verified journalists, attorneys, and platform representatives can reach me on LinkedIn for matters requiring direct response within five business days.

What I want from pitches. Primary-source access I do not have: regulatory filings, internal documents from platforms covered on the site, model-performer testimony, payment-processor source documents on chargeback patterns, academic research on harm and consent dynamics in AI companion products. What I do not want. Re-aggregated trend pieces, opinion essays without sourcing, AI-generated copy, and brand-side proposals dressed as story tips.

For fact-checking calls, methodology questions, or contestation of a published score: include the article URL, the specific claim being contested, and any supporting documentation. The contestation pathway is intentionally narrow. A brand can flag a factual error (I check and correct on the public errata board) or supply a documented product change and request an accelerated retest (I assess against the schedule and the trigger criteria). A brand cannot negotiate a score or buy a placement.

How is bestgirlfriend.ai different from competitors?

The competitive backdrop is the news angle. The most-cited adult-AI review sites in 2026 share three patterns: anonymous or fabricated editor bylines, a single "How we review" page that says "we test every product" without naming who tested it or when, and undated review pages with broken outbound links to expired offers. The well-resourced top-ranking competitor on "best AI girlfriend" search results in EN-US lists a "Coursera Bachelor 2005-2009" credential on its About page, which is a credential that cannot exist because Coursera was founded in 2012. Several other directories publish "expert reviews" with no editor named, no testing protocol disclosed, and rankings that change in lockstep with payout changes.

The absence of those three artifacts (named editor, published method, public corrections log) is the diagnostic. If a publication will not say who tested the product and when, the score is unfalsifiable. If the page has no dateModified, the score is undated. If there is no public errata log, there is no way to tell whether the publication corrects mistakes or pretends they never happened. The three positive references I built against on the trust side are Wirecutter's "How We Test", ProPublica's corrections policy, and Consumer Reports testing-and-disclosure standards. None of these institutions match in resource or reach; the editorial discipline ports anyway.

A small team can adopt the named-scoring, score-lock, and public-corrections pattern with no infrastructure heavier than what is documented here. That is the story underneath the press kit, and journalists are welcome to verify every claim with the links above.

Frequently asked questions

Cite this page

Joly, A. (2026). Press Kit: Bio, Brand Assets, Citation Format. bestgirlfriend.ai. https://bestgirlfriend.ai/press

Sources

• [Source: Alexandra Joly, LinkedIn (Senior Editor at bestgirlfriend.ai) · verified 2026-05-26], verifiable professional profile, education, work history.
• [Source: Alexandra Joly, Bluesky (DNS-verified handle alexandra.bestgirlfriend.ai) · verified 2026-05-26], cryptographic proof of domain control via TXT record.
• [Source: Université Sorbonne Nouvelle (Paris), public university accreditation · verified 2026-05-26], degree-issuing institution for Licence Information et Communication 2013-2016.
• [Source: Wirecutter, About How We Test · verified 2026-05-26], positive-anchor reference for named accountability and published methodology.
• [Source: ProPublica, Corrections Policy · verified 2026-05-26], positive-anchor reference for public corrections log.
• [Source: Consumer Reports (Wikipedia) · verified 2026-05-26], positive-anchor reference for non-profit independent testing.
• [Source: Coursera (Wikipedia) · verified 2026-05-26], founding year 2012 (verifies the competitor-credential-cannot-exist factual claim).
• [Source: FTC 16 CFR Part 255, Guides Concerning the Use of Endorsements and Testimonials in Advertising · verified 2026-05-26], federal basis for the affiliate disclosure layer.

Related pages

• About bestgirlfriend.ai and Alexandra Joly, full publication and editor bio (broader scope than this press kit).
• Editorial process, the 12-step workflow producing every review.
• Methodology landing page, the four scoring frameworks in one place.
• AI Companion Scoring, 8 categories with weights.
• Cam Site Scoring, 6 categories, $0-spend protocol.
• Adult Game Scoring, 7 categories with Billing Transparency.
• Real Models Scoring, 6 categories per-creator.
• Benchmark Dataset, 49 platforms scored, CC-BY 4.0, ready for citation.
• Affiliate Disclosure, commercial firewall and FTC 16 CFR Part 255 compliance.
• Contact, full mailbox routing with per-purpose service-level commitments.
• Errata board, public log of corrections.

---

Last verified May 26, 2026 · See errata log for any post-publish corrections · Editor: Alexandra Joly · Methodology · Editorial process · Affiliate disclosure