OpenAI извинились за новый дизайн ChatGPT

OpenAI признала, что новый десктопный ChatGPT вышел перегруженным: слитые в один интерфейс Chat, Work и Codex запутали пользователей и раздули приложение почти в 10 раз. После хейта обновление откатили, а вывод простой: функции надо упаковывать в единый UX без лишнего шума, даже если это помогает росту базы.

➡️ Читайте на сайте: https://aff.top/blog/openai-izvinilis-za-novyi-dizain-chatgpt

🧠 Ещё больше инсайтов → в канале AFF.top

Dated 'last reviewed' stamps on YMYL: a 300-page recency study

The question: does adding and maintaining a verifiable 'last reviewed' date — backed by actual content updates — correlate with YMYL performance, or is the date itself decorative?

A medical publisher's documented rollout: ~300 evergreen condition pages received a visible 'Last medically reviewed on [date]' stamp tied to a real quarterly review cycle, where reviewers logged changes. Pages genuinely updated (new studies cited, dosages corrected) were compared against pages where the stamp refreshed but content barely changed. Reported result over ~6 months: substantively-updated pages gained an average of 2.6 positions; stamp-only-refreshed pages gained ~0.3.

This maps to the QRG's emphasis on accurate, current information for YMYL (section 3.4) and to Google's repeated warnings against fake freshness.

Counter-evidence: substantively-updated pages received editorial attention the stamp-only pages did not — again, effort confound.

Caveat: the date string itself is almost certainly not the lever. The case suggests the *content change behind* the date matters, and a refreshed date over stale content moved nothing.

What we still don't know: whether Google detects 'fake' freshness directly, or whether stamp-only pages simply failed to improve because nothing about them actually improved. Either reading fits the data.

ChatGPT 5.6 Luna и Terra подешевели

OpenAI резко снизила цены на GPT-5.6 Luna и Terra спустя три недели после релиза, сделав их заметно дешевле для массового использования. Sol осталась премиальной по прежнему прайсу, но получила Fast mode с ускорением до 2,5 раза. Вывод: компания давит ценой и скоростью, чтобы быстрее нарастить спрос и долю рынка.

➡️ Читайте на сайте: https://aff.top/blog/chatgpt-5-6-luna-i-terra-podesheveli

🧠 Ещё больше инсайтов → в канале AFF.top

Доменная зона .web делегирован в корневую зону DNS

Verisign вывела .web в корневую зону DNS: теперь это полноценный TLD, но открытая регистрация ещё не стартовала. Сначала доступ получат владельцы совпадающих доменов в .com через LRP. Вывод для рынка: хорошие EMD в .com могут стать входным билетом в .web, если успеть занять брендовые имена раньше общего запуска.

➡️ Читайте на сайте: https://aff.top/blog/domennaia-zona-web-delegirovan-v-kornevuiu-zonu-dns

🧠 Ещё больше инсайтов → в канале AFF.top

В Facebook Ads появился раздел «Conversations»

Facebook Ads добавил Conversations с автоответом на комментарии: по ключевым словам можно сразу отправлять сообщение в личку. Для арбитража это новый способ прогрева и передачи ссылки без клоаки: в креативе можно просить оставить комментарий, а заинтересованных уводить с вайта на блэк уже в ДМ. Идея спорная, но её стоит тестировать.

➡️ Читайте на сайте: https://aff.top/blog/v-facebook-ads-poiavilsia-razdel-conversations

🧠 Ещё больше инсайтов → в канале AFF.top

Earning trade-press citations: a 12-month authority-building study

The question: do earned third-party citations — being quoted and linked by established outlets — correlate with topical ranking better than on-page EEAT changes?

A SaaS company's documented program: over 12 months it pursued earned mentions, landing ~28 citations across recognized trade and mainstream outlets (quotes in articles, original-data citations, expert roundups). No major on-page EEAT overhaul ran in parallel. Reported outcome: the company's resource hub moved from page-two averages to top-five for ~45 mid-competition queries, and brand-entity searches rose ~60%.

This is the QRG's reputation construct in its purest form (section 2.6): 'what others say' about the site and its creators, sourced independently. It is also the closest thing to a clean test, because on-page was held roughly constant.

Counter-evidence: many of those citations carried links, so classic link equity confounds the EEAT interpretation. We cannot cleanly separate 'reputation signal' from 'backlink.'

Caveat: 28 citations over a year is also 28 PR-quality content efforts, each of which may have produced spin-off coverage and social signal.

What we still don't know: how much of the lift is reputation-as-EEAT versus plain link authority. The two are deeply confounded in every real-world citation, which is precisely why on-page-only EEAT tests rarely move much.

A 4,000-page AI-content site and its authority collapse

The question: at scale, does mass-produced content lacking experience signals fail specifically on trust, or does it fail for ordinary quality reasons?

The widely-discussed 2023 case: a programmatic site published ~4,000 AI-written articles, briefly ranked across thousands of queries, peaked around an estimated 200,000+ monthly clicks, then lost the overwhelming majority of traffic across subsequent core and helpful-content updates. The content had no named authors with verifiable expertise, no first-hand experience, and no off-site reputation.

The failure is consistent with HCU's 'people-first' framing and the QRG's instruction that content created with little effort and no demonstrable expertise on YMYL-adjacent topics warrants low ratings.

Counter-evidence: the site also stuffed keywords, had a brand-new domain, and grew unnaturally fast — any of which could trigger demotion independent of EEAT.

Caveat: the case proves a site *can* collapse, not *why*. 'No experience signals' and 'low effort, fast scaling, thin domain trust' are bundled together. AI authorship per se is not isolated as the cause.

What we still don't know: whether the same content under a named expert with real off-site reputation would have survived — which is the actually-interesting question, and one no public case has cleanly answered.