Maurício de Queiroz

Construindo a @trybases. Pesquisando como as buscas com IA impactam as marcas.

[Maurício de Queiroz @itsmqueiroz]

Estamos construindo uma nova geração de tecnologia para profissionais de marketing.

Acreditamos que a disciplina atravessa uma das maiores transformações de sua história.

[Maurício de Queiroz @itsmqueiroz]

Ontem a OpenAI abriu em beta o Businesses in ChatGPT. A marca reivindica o perfil, ajusta as informações e vê como as pessoas interagem com ela.

O detalhe que mais me chamou atenção: o painel só conta "menção" quando o ChatGPT cita o site da empresa. Falar só o nome da marca não conta.

Na prática, o domínio virou a identidade da marca dentro do ChatGPT. Quem tem nome ambíguo e site mal amarrado a ele vai parecer menor no painel do que realmente é nas respostas.

O primeiro dado oficial do ChatGPT já vem com esse recorte.

[Maurício de Queiroz @itsmqueiroz]

@ThamiBenjelloun One of the most honest Search Console screenshots I've seen this week. That page shows up in AI answers because it answers a concrete question. The hard part is making the answer lead to deliverability without turning a useful page into a sales page that stops being cited.

[Maurício de Queiroz @itsmqueiroz]

@ajavuu True, but brand spend lived with this for decades, and the answer was never perfect attribution. It was leading indicators the exec team agreed on upfront. The bigger risk: a competitor becoming the default recommendation in your category while you wait for a clean revenue line.

[Maurício de Queiroz @itsmqueiroz]

@KateBour The packaging part now has a second audience. If you can't explain what you do in one clear sentence, an AI assistant will write that sentence for you, from whatever it found. Usually worse.

[Maurício de Queiroz @itsmqueiroz]

@KateBour Agree, with one shift: when content gets cheap, the scarce asset isn't content. It's being the name people, and now models, already associate with the problem. Good marketing used to win attention. Now it also has to win the default answer.

[Maurício de Queiroz @itsmqueiroz]

@AravSrinivas Same law applies to brands in AI answers: left alone, the entropy of how a model describes you only goes up. Somebody has to keep the source of truth maintained.

[Maurício de Queiroz @itsmqueiroz]

@CyrusShepard At this point Google Flights isn't competing in search, it is the search. AI answers are running the same play: the platform keeps the intent, and everyone else fights for the citation.

[Maurício de Queiroz @itsmqueiroz]

@rustybrick The interesting part isn't the scraping, it's who depends on it. A lot of AI answer products are built on top of someone else's SERP. Every ruling like this is quietly a ruling on how answer engines get to source the web.

[Maurício de Queiroz @itsmqueiroz]

@Marie_Haynes The real risk is misreading it: stores blaming the update for a Black Friday drop, or crediting Black Friday for an update gain. Two big signals in the same window. I'd annotate everything now, before the dashboards turn into opinions.

[Maurício de Queiroz @itsmqueiroz]

@rustybrick Too early for rankings on my side, but I'm watching something else: whether AI Overview citations shift with the update or lag behind it. If they move on a different clock than the blue links, that tells us a lot about how separate those systems really are.

[Maurício de Queiroz @itsmqueiroz]

@RossHudgens Aggregation cuts both ways. It weakens the one random anonymous review, but it also turns the aggregated verdict into a single point of failure. Brands stop fighting individual comments and start fighting a summary.

[Maurício de Queiroz @itsmqueiroz]

@lilyraynyc Same reading here, with a side effect: the dashboard measures recommendations, not reputation. A brand can be described in a lot of answers without a link and show zero here. The gap between 'talked about' and 'linked' is probably the most interesting number nobody sees yet.

[Maurício de Queiroz @itsmqueiroz]

@Kevin_Indig The detail I'd watch is the one you flagged: mentions only count when the website is referenced. That makes the domain the entity key. Brands with an ambiguous name and a site that is poorly tied to it will look smaller in this dashboard than they really are in the answers.

[Maurício de Queiroz @itsmqueiroz]

The detail I'd watch is the one you flagged: mentions only count when the website is referenced. That quietly makes the domain the entity key. Brands with an ambiguous name and a site that is poorly tied to it will look smaller in this dashboard than they really are in the answers.

[OpenAI @OpenAI]

Now in beta: Businesses in ChatGPT.

Claim your business, manage your information, and get insights on how people engage with you.

Eligible businesses can also build a Business Agent to answer customers’ questions and help them explore what you offer.

https://t.co/N2GEZX0mFT

[Maurício de Queiroz @itsmqueiroz]

@simonredux @trybases

[Maurício de Queiroz @itsmqueiroz]

@Marie_Haynes I'd treat this as an Ultra rollout for now, not a site-wide update.

If intent handling changes, I'd watch what the answer says, not only where the page ranks.

[Maurício de Queiroz @itsmqueiroz]

@chris_nectiv The split is the finding I'd keep.

A single average for the market would have hidden the sites that barely moved.

[Maurício de Queiroz @itsmqueiroz]

Uma captura de um único modelo não descreve a marca.

A mesma pergunta muda de um lugar para outro. Por isso faz sentido olhar a visibilidade por plataforma, não só o que aparece numa janela.

[Elon Musk @elonmusk]

We have dumped JAX/SGLang and rewritten the entire RL and inference stack from scratch in C/C++, which took some extra time.

Training (aka pre-training) is done. Lot of cleanup was needed on the SFT data to get that mostly right. Now, 4.8 is in RL and the evals are starting to climb. We will do a soft rollout to Bot users later this week and see how it goes.

Grok 4.8 will bear no resemblance to 4.7 (thankfully). It is not yet incredible, but, this time, the foundations are built right, so I expect it to evolve to be something special.