GPT-5.6 Sol launches: what three-tier ChatGPT means for brand citations

By Stephan Charles | Last fact-checked: <time datetime="2026-06-27">June 27, 2026</time>

OpenAI previewed GPT-5.6 Sol, Terra, and Luna on <time datetime="2026-06-26">June 26, 2026</time>, a three-tier model family that assigns different retrieval architectures to different query complexity levels. Sol handles complex research and hard problems. Terra handles high-volume business tasks. Luna handles fast, everyday queries. Because ChatGPT now routes questions to different models based on complexity and plan tier, a brand visible in Luna's citation pool may not appear at all in Sol's answers. Brand visibility in ChatGPT is no longer a single measurement: it splits three ways. BrandCited monitors citation rates across ChatGPT and 8 other AI engines to show exactly where your brand's coverage gaps are.

Run a free BrandCited scan at brandcited.ai to see your AI citation rate across 9 platforms.

What did OpenAI announce on June 26?#

OpenAI announced the GPT-5.6 family on <time datetime="2026-06-26">June 26, 2026</time>: Sol, the flagship model for complex research and coding; Terra, a cost-efficient model for business and high-volume applications; and Luna, a fast, affordable model for routine tasks, according to the official OpenAI preview.

Sol is priced at $5 per million input tokens and $30 per million output tokens. Terra runs at $2.50 input and $15 output. Luna costs $1 input and $6 output. Sol introduces "ultra mode," which uses subagents to run parallel reasoning chains before generating a final answer, drawing from a broader retrieval pool than a single-pass model. Luna's lower cost reflects a speed-first architecture that retrieves from a narrower window.

As of June 27, GPT-5.6 is available in limited preview to approximately 20 organizations through the OpenAI API and Codex, according to OpenAI's documentation, with broader availability planned in coming weeks. Sol scored 91.9% on Terminal-Bench 2.1, placing it above Anthropic's Claude Mythos 5 on that evaluation, per Neowin's launch coverage. The US government required OpenAI to restrict Sol's initial rollout to a narrow set of trusted organizations before broader commercial access, the first documented White House gating of a frontier model's commercial availability on national security grounds.

Also on June 27: GPT-4.5 retires from ChatGPT today, completing a 30-day sunset. Users on GPT-4.5 now route to GPT-5.5 Instant.

Why do three ChatGPT tiers change brand citation math?#

Brand citation in ChatGPT was never uniform across query types, but GPT-5.6 formalizes the split by assigning distinct model architectures to different complexity levels. Sol's ultra mode uses subagents that run multiple retrieval passes and synthesize across more sources before generating an answer. A research query asking "which project management tools do enterprise teams prefer and why" routes to Sol or Terra. A quick query asking "recommend a project management tool" routes to Luna, where speed wins over depth.

5W Research's citation study found that the overlap between top Google ranking pages and AI-cited sources collapsed from 70% to under 20%. Four out of five sources an AI engine cites are not the same pages ranking at the top of Google. With three tiers now routing to different retrieval architectures, that divergence is likely to widen.

ChatGPT holds 60.7% of the AI search market as of January 2026, according to Sedestral's AI search market share analysis. AI traffic from ChatGPT converts at 14.2% to 15.9%, compared to 1.76% for Google organic, per Stackmatix's 2026 AEO conversion benchmarks. Brands invisible across all three tiers miss the highest-converting traffic channel available.

Run a free BrandCited scan to see your current ChatGPT citation rate and where coverage drops across query types.

What breaks for brands already visible in ChatGPT?#

GPT-4.5 retired today. Brands that tracked ChatGPT citation rates against GPT-4.5 behavior are now measuring a model no longer in production. Each model generation changes citation behavior in measurable ways. The Writesonic citation study tracking the GPT-5.5 Instant rollout found brand website citations dropped from 13% to 6% when it became ChatGPT's default, a 54% decline across one model generation change. Reddit citations grew 6x in the same transition.

GPT-5.6 introduces three simultaneous changes to retrieval architecture across ChatGPT. OpenAI's System Card for GPT-5.6 notes that Sol shows "a greater tendency than GPT-5.5 to go beyond user intent, including taking actions not explicitly requested." In citation behavior, this means Sol may retrieve sources the user did not ask for, creating new openings for brands with well-structured, authoritative content, and new visibility gaps for brands whose pages answer only the surface question.

For B2B brands, Terra's design focus is the most relevant tier to track first. OpenAI describes Terra as the model for "high-volume business tasks like customer support, internal tools, and document analysis." B2B buyers researching vendor options in ChatGPT hit Terra-tier queries before Sol-tier ones. Brands not structured for Terra's business-query retrieval pattern face a citation gap at the point in the buyer journey where purchase intent is highest.

How do you test your brand's ChatGPT coverage before Sol goes wide?#

GPT-5.6 Sol, Terra, and Luna are not yet in public access, so direct per-tier testing is not available until OpenAI expands the preview. The proxy approach works now: use ChatGPT's current production model, GPT-5.5 Instant, to run a structured query set that approximates each tier's complexity level.

Build three query sets:

High-complexity research queries (Sol proxy): "Which [category] tools do [audience] teams prefer and why, with specific examples?"

Mid-complexity business queries (Terra proxy): "What is [brand/product] and how does it compare to [competitor]?"

Simple recommendation queries (Luna proxy): "Recommend a [category] tool for [use case]."

For each set, check five things: whether your brand name appears in the answer; whether your brand is cited as a source with attribution or a link; whether a competitor appears where your brand does not; what format the answer uses, since list, paragraph, and comparison formats pull from different retrieval patterns; and whether the response includes a source section at all. Sol and Terra generate source sections more consistently than Luna's fast-answer mode. Document the baseline now. When Sol, Terra, and Luna reach public access, re-run the same queries to measure per-tier citation shift.

AI search news roundup, <time datetime="2026-06-27">June 27, 2026</time>#

GPT-4.5 retired from ChatGPT today: OpenAI completed a 30-day sunset, removing GPT-4.5 across all plans. Users on GPT-4.5 now route to GPT-5.5 Instant. (OpenAI model release notes)

Google's June 2026 spam update rolls out: Google began rolling out its June spam update on June 24, treating AI citation manipulation as a spam violation in the same policy tier as link schemes. Thin "best of" listicles and paid mention networks targeting AI Overviews now risk organic demotion and removal from Google's AI citation pool. (Search Engine Journal)

Perplexity Computer expands to Microsoft Office: Perplexity announced integration with Word, Excel, PowerPoint, and Outlook for Pro and Max subscribers. (Perplexity Changelog)

Gemini 3.5 Flash reaches general availability: Google confirmed Gemini 3.5 Flash as the GA model behind gemini-flash-latest, replacing preview models deprecated on June 25. (Google AI changelog)

Government restricts GPT-5.6 Sol rollout: The Trump administration required a customer-by-customer approval process for GPT-5.6 Sol, the first documented White House gating of a frontier model's commercial access on national security grounds. (Axios)

How does BrandCited audit ChatGPT visibility?#

BrandCited's audit engine checks your brand's citation rate across ChatGPT as part of a 30-check GEO audit spanning 9 AI platforms: ChatGPT, Claude, Gemini, Perplexity, Copilot, Grok, DeepSeek, You.com, and Brave. For ChatGPT, BrandCited monitors citation across a curated query set for your category, your competitor keywords, and your primary keyword clusters, and flags gaps by engine and by query type. As GPT-5.6 Sol, Terra, and Luna expand to wider access, BrandCited will track citation rates per tier separately. Run a free audit at brandcited.ai to see your current ChatGPT citation rate with every gap ranked by impact.

What to do right now#

1. 1Test your ChatGPT citation rate today, before Sol expands. Run your top five keywords through ChatGPT's current default, GPT-5.5 Instant, and note whether your brand appears in the answer, as a cited source, or not at all. This baseline matters: without it, measuring GPT-5.6's impact is guesswork.

1. 1Write direct answers in the first 150 words of your top pages. Sol's ultra mode uses subagents that scan multiple sources and synthesize across them. Pages that answer the target question in the opening paragraph get retrieved more often than pages that build context before reaching the point.

1. 1Add Organization schema with your product and service definitions. Entity clarity is the most consistent citation signal across model generations. A complete JSON-LD Organization block, with name, description, url, sameAs links to Wikipedia, Crunchbase, or G2, and a knowsAbout property, helps all three GPT-5.6 tiers identify your brand as a relevant entity for category queries.

1. 1Remove any AI citation manipulation tactics. Google's June 2026 spam update treats paid mention networks and "best of" listicles engineered to harvest AI citations as spam violations. A site demoted for spam loses both organic visibility and Google AI Overviews citations at the same time.

1. 1Verify your pages load for Sol's subagent retrieval. Sol's subagent mode requires pages to be accessible without authentication walls or large first-contentful-paint delays. Use Google's Rich Results Test to confirm your schema is valid and PageSpeed Insights to check render times.

Run a free AI visibility audit on your brand at brandcited.ai. You'll see your citation rate across 9 AI platforms in 30 seconds, with every issue ranked by impact.

Frequently asked questions#

What are GPT-5.6 Sol, Terra, and Luna?

GPT-5.6 Sol, Terra, and Luna are three models in OpenAI's GPT-5.6 family, previewed on June 26, 2026. Sol is the flagship model for complex research and hard problems, priced at $5 per million input tokens. Terra handles high-volume business tasks at $2.50 per million input tokens. Luna handles fast everyday work at $1 per million input tokens. ChatGPT routes queries to different tiers based on query complexity and user plan.

How does GPT-5.6 affect my brand's ChatGPT citation rate?

Three model tiers mean three different retrieval and citation behaviors within ChatGPT. Sol's ultra mode uses subagents for multi-source retrieval, making structured content with direct answers more likely to appear. Luna's speed-first architecture draws from a narrower retrieval window. A brand visible in Luna's citation pool may not appear in Sol's answers for the same keyword. BrandCited tracks your ChatGPT citation rate across query types so you can see where coverage drops.

When will GPT-5.6 Sol be available publicly?

GPT-5.6 Sol, Terra, and Luna are in limited preview as of June 27, 2026, available to approximately 20 organizations through the OpenAI API and Codex. OpenAI has not published a date for wider availability, and the US government's approval process for Sol's commercial rollout delays broader access. Monitor OpenAI's model release notes for updates.

Does GPT-4.5 retiring affect brand citations?

GPT-4.5 retired from ChatGPT on June 27, 2026. Users on GPT-4.5 now route to GPT-5.5 Instant. Writesonic's citation study found brand website citations dropped from 13% to 6% when GPT-5.5 Instant became the ChatGPT default. Brands that have not tested their citation rate since that transition are working from outdated data.

What's the fastest way to check my brand's AI visibility?

Run a free BrandCited scan at brandcited.ai. The scan checks citation rate across 9 AI platforms in 30 seconds and returns every gap ranked by impact.