GPT-6 Astra browses like a person: what ChatGPT's new engine means for brand citations

GPT-6 Astra does not scrape static text from web pages. It opens pages, clicks navigation links, handles popups, and fills forms, running tasks across multiple tabs in sessions lasting up to 40 minutes, according to OpenAI's product overview. The browsing pattern matches how a researcher navigates a site, not how a crawler indexes it.

For paid ChatGPT subscribers, Astra is now the model generating cited answers in ChatGPT Search. When a user asks "what's the best project management tool for remote teams," Astra visits pages it considers candidates, reads them at depth, and decides which brand earns the citation.

Why does Astra's human-style browsing change citation decisions?#

Astra changes citation decisions because it evaluates page depth and clarity the way a person does, not the way a keyword-matching crawler does. A Mintec analysis published after the Astra launch found that Astra reads pages at depth and decides whether a brand deserves the citation, rather than skimming search snippets.

The Semrush 2026 AI Visibility Index, which analyzed 126 million AI search prompts from January through April 2026, showed that only 11% of domains are cited by both ChatGPT and Perplexity, meaning the two engines apply different criteria for source selection. ChatGPT cites an average of 15 sources per response, compared to Gemini's average of 3 sources per response, per the same index.

Wikipedia captures roughly 47.9% of ChatGPT's top citations, according to ContextBolt's analysis of Astra citation patterns. Wikipedia earns those citations because every article states its key facts in the first sentence of each section. That is the same content structure Astra rewards in brand content.

Which signals does Astra prioritize when it browses your site?#

Clear entity definition, real third-party mentions, and structured pricing or service information are the three site signals Astra's human-style browsing rewards, according to ContextBolt's GPT-6 Astra SEO analysis. The signals are not new to GEO. What changed is how thoroughly Astra can verify them during a session.

Brand citation rates differ across platforms by an order of magnitude: Perplexity cites brands at 13.05% compared to ChatGPT's 0.59%, per the Airops 2026 State of AI Search report cited in QuickSEO's AI visibility comparison. The gap does not mean ChatGPT cites fewer strong sources. It reflects ChatGPT's higher confidence threshold before it names a brand in an answer.

Brands cited inside AI Overviews earn 35% more organic clicks and 91% more paid clicks than non-cited brands on the same SERP, according to data from SQ Magazine's AI Overviews statistics roundup. Astra's human-style browsing means the brands earning those citation spots are the ones with content that communicates trust to a reader, not just to a crawler.

What breaks when brands optimize for crawler snippets instead of human browsers?#

Ghost citations happen: an AI engine cites your page as a source but never mentions your brand in the answer text. Kevin Indig and Semrush identified this pattern in their June 2026 ghost citations study, which found that 62% of AI citations across four major engines are ghost citations where the brand name never appears in the answer.

The split between citation rate and mention rate varies by engine. On ChatGPT, the citation rate reaches 87% but the brand mention rate is only 20.7%. On Gemini, the mention rate is 83.7% but the citation rate is only 21.4%, per the same Semrush study. A brand optimized only for getting its URL into a source list ends up invisible to the reader even when it wins the citation slot.

BrandCited tracks both citation rate and mention rate per engine. Run the free scan at brandcited.ai to see where your brand is cited but not named, and which pages need a clearer lead sentence.

What else changed in AI search this week?#

These are the AI search developments from September 28-29, 2026 that affect brand citation strategy.

• Gemini in Chrome address bar: Google began rolling out Gemini AI Mode in the Chrome address bar for US users in English by the end of September 2026, making AI-generated answers accessible directly from search for a broader user base. (The Neuron, September 28, 2026)

• Perplexity Effort Mode: Perplexity released Effort Mode with four levels: Light, Standard, High, and Ultra. Ultra mode runs more sources and spends more compute per query. Brands with deeper, more structured content have a path to capture citations at the Ultra level that lightweight content cannot reach. (Perplexity release notes, September 2026)

• Google AI Overviews claim fidelity data: An academic study covering 55,393 queries found that 11.0% of 98,020 atomic AI Overview claims are unsupported by the pages Google cites, with omission as the dominant failure mode. (arxiv.org study, 2026)

• Google AI Mode sponsored ads testing: Google was spotted testing text link ads inside AI Mode, with sponsored anchor text woven into answers, with a "Sponsored" label visible to users. The test, which started September 15, signals that AI-generated answers will carry paid placements for some queries. (Silverback Marketing, September 2026)

• Perplexity and HP partnership: HP announced it will pre-install Perplexity on Windows PCs starting with the HP ZBook Ultra G3a in October 2026. The partnership extends Perplexity's reach into enterprise hardware, adding a new surface where brand citations can generate referral traffic. (Perplexity release notes, September 2026)

How does BrandCited audit Astra-style citation readiness?#

BrandCited's audit engine runs 30+ checks against each brand's web presence, including the signals that Astra's human-style browsing evaluates. The entity-clarity check verifies that each page defines what your brand does, who it serves, and what category it belongs to within the first 150 words. The content-structure check flags pages where the key claim is buried in paragraph four instead of leading the first sentence. The ghost-citation check shows where your domain is cited but your brand name is absent from the answer text. Run the free audit at brandcited.ai to see your citation readiness score across nine AI engines in 30 seconds.

See your brand's AI citation score across nine engines, including ChatGPT now powered by GPT-6 Astra. BrandCited's audit identifies every citation gap and ranks fixes by impact. Free scan at brandcited.ai. No signup required.

What to do right now#

1. 1Check your entity definition on every key page. Open your homepage and top three landing pages. The first paragraph of each should name exactly what your brand does, who it does it for, and what category it occupies. If the definition is vague or buried, Astra cannot place your brand in a relevant answer with confidence.

1. 1Lead every H2 section with the answer, not the setup. The first sentence of every section should answer the question the heading poses. Astra rewards pages that answer before they explain. Wikipedia, the source capturing 47.9% of ChatGPT's top citations, opens every section with the fact, not the context.

1. 1Add Article and FAQPage schema to your top content. Structured data lets Astra extract author, date, and Q&A facts without parsing every paragraph. BrandCited's scan data from 847 brands shows that sites with complete Article and FAQPage schema score 23 points higher on the composite AI visibility index than comparable sites without schema.

1. 1Check for ghost citations in your current AI citation data. If your brand earns citations but not mentions, you are losing recognition even when you win the source slot. BrandCited tracks both citation rate and mention rate per engine, so you can see exactly where the gap exists and which pages need a stronger brand name in their lead content.

1. 1Monitor ChatGPT Search for Astra-era citation shifts. The citation logic Astra applies will stabilize over the coming weeks. BrandCited runs weekly citation probes across ChatGPT, Perplexity, Claude, Gemini, Copilot, Grok, You.com, and Brave Search. Run the first scan now to establish a baseline before Astra's patterns settle.

FAQ#

What is GPT-6 Astra and how does it differ from previous ChatGPT models?

GPT-6 Astra is OpenAI's latest model, reaching stable release on September 25, 2026. It differs from earlier models in its agentic browsing capability: instead of relying on scraping static text, Astra navigates websites like a person, clicking links, filling forms, and running multi-tab sessions lasting up to 40 minutes. For brand citations, the content quality a human reader experiences now determines whether Astra cites your brand.

Does GPT-6 Astra affect all ChatGPT users or only paid subscribers?

GPT-6 Astra powers ChatGPT Search for paid subscribers on Plus and Pro plans. Free-tier ChatGPT users interact with earlier model versions for standard chat queries. Because most high-intent product research queries run through ChatGPT Search, the paid user segment is the citation surface that matters most for brands optimizing for AI visibility.

What is a ghost citation and why does it matter for brand visibility?

A ghost citation occurs when an AI engine cites your page as a source in its answer but never mentions your brand name in the answer text. The Semrush and Kevin Indig ghost citations study from June 2026 found that 62% of AI citations across four engines are ghost citations. A ghost citation drives zero brand recognition even when the AI considered your content credible enough to use as a source.

How is ChatGPT's citation behavior different from Perplexity's?

ChatGPT cites an average of 15 sources per response and has a brand citation rate of 0.59%. Perplexity has a brand citation rate of 13.05%. On ChatGPT, citation rates reach 87% but brand mention rates are only 20.7%, making ghost citations a larger problem there than on Perplexity, where brands that are cited are named 83.7% of the time. Platform-specific optimization strategies matter because the two engines use different citation logic.

How can BrandCited help with GPT-6 Astra citation readiness?

BrandCited runs citation probes across nine AI engines, including ChatGPT powered by GPT-6 Astra, and scores each brand on citation frequency, citation prominence, URL attribution rate, and cross-engine consistency. The audit flags pages with weak entity definition, missing schema, and content structure that buries the answer. The free scan at brandcited.ai takes 30 seconds and requires no signup.