Why Your Brand Doesn't Show Up in Comet or Atlas

Perplexity Comet went generally available on iOS, Android, Mac, and Windows simultaneously in May 2026, the first AI browser to ship cross-platform parity. ChatGPT Atlas crossed 10 million active users in its first 90 days. Dia from The Browser Company is bundled inside Arc successor installs. Brave Leo ships with every Brave install. Microsoft Edge Copilot sidebar is on by default for new Windows 11 setups.

The AI browser is now a third citation surface, distinct from chatbots and from AI Overviews. Most B2B SaaS brands have never tested whether their pages appear inside one. The few that have run the test usually find their brand absent in every prompt the sidebar agent answers.

The gap is not a search ranking problem. The retrieval mechanics are different in ways that punish brands optimized only for SERP-style indexing.

AI browsers do not search. They synthesize from what the user is already looking at.

If your page is not in the active tab, in the recent history, or pulled into the sidebar by a click, the agent treats your brand as if it does not exist. That is true even when your domain ranks first in Google.

What an AI browser actually is

The category is six products old in May 2026 and still hardening. Each one routes citations through different pools, and none of them behave like ChatGPT search.

Pattern #1: The sidebar agent reads the active tab first

Comet, Atlas, Dia, and Edge Copilot all prioritize the page rendered in the open tab when the user asks a question. The agent extracts the visible DOM, summarizes it, and uses it as the dominant context for any follow-up answer. If a competitor's pricing page is open and the user asks "is this the right plan for my team," the competitor wins the citation by default.

Pattern #2: The session history is the secondary corpus

When the active tab cannot answer the question, the agent falls back to pages the user opened earlier in the session. Comet's documentation explains that the agent "reasons across tabs you have visited recently." Atlas calls the same pool "browsing context." Dia routes through "tab memory." In every product, the pages a user already clicked have priority over the open web.

Pattern #3: The web index is the tertiary fallback

Only when the active tab and the session history both fail does the agent fall back to a web search index. Comet uses Perplexity's own retrieval stack. Atlas routes to ChatGPT search, which uses Bing as its primary index. Edge Copilot uses Bing directly. The tertiary fallback looks like classical AI search, but it fires roughly a third of the time, not most of the time.

Pattern #4: The bookmark and saved-page pool gets weighted

Comet and Dia both surface the user's bookmarked pages as candidate citations for relevant queries. A user who has bookmarked your product documentation is far more likely to see your brand cited inside that user's AI browser than a user who has never visited.

Why your brand is invisible in AI browsers

Most B2B brands have never run the test. The ones that have run it find their citation rate inside AI browsers is below 5% even when their organic ranking is strong. The gap traces to five structural reasons.

Reason #1: Your buyer never opens your page in the same session

Buyers research alternatives. A finance-ops leader evaluating spend management opens three tabs from three vendors, asks the sidebar agent to compare them, and gets back a synthesis built from the three open tabs. The vendor whose tab was never opened is excluded from the entire answer. If you do not show up in the buyer's first three Google clicks, you do not show up in the AI browser synthesis.

Reason #2: Your page renders too slowly for the agent to extract

The sidebar agent waits a fixed time window for the DOM to finish rendering before extracting. Comet's window is around 2.5 seconds. Atlas extends slightly longer but still cuts off. Pages that depend on heavy client-side hydration, deferred above-the-fold images, or third-party scripts often miss the window. The user sees the page, but the agent sees a half-finished extract and downweights the source.

Reason #3: Your key claims are buried below the fold

The agent extracts visible content first and scrolls only when the user does. A pricing page that shows a hero animation in the first viewport and the actual pricing on scroll gives the agent nothing usable. The buyer asks "how much does this cost" and the agent answers "I cannot tell from this page," which is a worse citation outcome than not appearing at all.

Reason #4: Your content is not in the buyer's bookmark pool

Brands that operate documentation, integration directories, or implementation guides that buyers actually save and re-open earn a structural advantage inside AI browsers. Brands that operate only marketing pages get visited once and never re-opened. The marketing page is invisible to the agent on every subsequent session.

Reason #5: Your sources never make it into the tertiary index either

When the active tab and history both fail, the agent falls back to web retrieval. The retrieval pool is dominated by Wikipedia, Reddit, YouTube, trade press, and a small set of high-authority domains. If your brand is absent from those pools, the tertiary fallback also produces zero citations. The agent then synthesizes an answer without you in it.

How AI browsers differ from chatbots and AI Overviews

The retrieval model is different enough that an SEO-optimized page can rank first in Google, get cited regularly inside ChatGPT, and still be invisible in Atlas. The table below traces the contrast.

Surface	Primary input	Secondary input	What gets cited
Chatbot (ChatGPT, Claude)	Web retrieval pool	Training corpus	Top retrieval candidates re-ranked at synthesis
AI Overviews (Google)	Indexed SERP results for the query	Knowledge Graph	Top-ranking pages with schema and E-E-A-T signals
AI browser (Comet, Atlas, Dia)	Open tab and session history	Bookmark pool, then web search	Whatever the user is already looking at

The third row is the one most B2B brands underestimate. Comet and Atlas put a hard ceiling on how many citations the brand can win without first being clicked through traditional channels.

Ranking in Google gets you into the AI browser pool. Getting opened in the buyer's actual session is what gets you cited.

How to audit your AI browser citation rate

The audit takes one analyst about four hours per product. Six steps, in order.

Step 1: Build a 30-prompt evaluation set tied to buyer intent

Pull the same prompts you already use for LLM tracking. Add ten prompts that buyers ask inside the browser specifically: "is this the right plan," "how does this compare to [alternative]," "can this integrate with [tool we already use]." Buyer-context prompts produce different citation behavior than category-context prompts.

Step 2: Run each prompt three times in each AI browser

Run the set in Comet, Atlas, Dia, Edge Copilot, and Brave Leo with cleared session state. Run again with your own pages open as the active tab. Run again with three competitor pages open. The three runs reveal cold-start behavior, owned-page priority, and competitor-context bias.

Step 3: Record the cited sources at the URL level

Log every domain and every page the agent surfaces. The URL-level citation tracking workflow we use for ChatGPT and Claude applies here with one extra column: which input mode produced the citation (active tab, session history, or web fallback).

Step 4: Diagnose the missing-source pattern

If your brand never appears in the cold-start run, the gap is in the tertiary web index. If your brand appears only when your tab is open, the gap is in bookmark pool and session history. If your brand appears in active-tab runs but the cited claim is wrong, the gap is in DOM extractability. Each diagnosis points to a different fix.

Step 5: Prioritize the fix queue against effort

DOM extraction problems are cheapest to fix and produce the fastest citation lift. Bookmark-pool gaps require content that buyers actually return to. Tertiary index gaps require the full earned-media and Wikipedia program we cover in the brand authority playbook.

Step 6: Re-run the audit quarterly

AI browser citation behavior changes faster than chatbot citation behavior. Comet pushes weekly updates. Atlas pushes biweekly. Edge Copilot updates with every Windows feature drop. A citation share that holds in May can shift by August without any content changes on your side.

What the fix looks like in practice

Three layers of work, ordered by how quickly they show up in re-audit numbers.

Layer 1: Make your page extractable inside 2.5 seconds

Move the actual answer above the fold. Strip the hero animation. Inline the critical claims as plain text inside the first viewport. Render pricing, comparison facts, and product capabilities in HTML rather than client-side JavaScript. The HTML parity audit framework applies directly to the AI browser DOM extractor, since the agent is reading the rendered DOM in real time and a half-loaded page reads as half a source.

Layer 2: Build content buyers return to

Documentation, changelogs, integration directories, implementation guides, and pricing calculators are the content types buyers actually bookmark. Marketing pages and case studies are not. Move budget from one bucket to the other and the bookmark-pool surface area widens within a quarter.

Layer 3: Earn the tertiary index citations

Wikipedia, Reddit, YouTube, and trade press are the same pool that backs ChatGPT and Atlas's web fallback. The Wikipedia playbook, the Reddit and YouTube social citation work, and the trade-press earned-media program all carry over directly. AI browsers do not have a separate citation economy. They have a stricter one.

What this changes for your AEO program

Three implications worth budgeting against in the next planning cycle.

Click-through is suddenly load-bearing again

For four years, search marketers have watched click-through rates fall as zero-click answers ate the SERP. AI browsers reverse that pressure for one specific reason: the AI browser only cites what the user actually opened. Brands that win the click into the buyer's session get cited inside the AI browser. Brands that win zero-click visibility in AI Overviews get nothing inside AI browsers. The two surfaces are now optimizing for opposite outcomes.

Documentation outranks marketing for AI browser citations

The pages buyers re-open are documentation and implementation guides. The pages they open once and close are marketing. The bookmark and session-history pools heavily favor the former. Brands that have starved their documentation budget will find that gap shows up first inside AI browsers.

The audit cadence has to be quarterly, not annual

Comet ships changes weekly. Atlas ships changes biweekly. A brand that audits AI visibility annually will be tracking five-month-old behavior on the AI browser side. Quarterly cadence is the floor, and monthly is reasonable for brands with material exposure to AI browser citations.

FAQ

How many B2B buyers actually use AI browsers in May 2026?

Atlas crossed 10 million monthly active users in its first 90 days, weighted heavily toward technical and product roles. Comet's user base is smaller but skews further into B2B research workflows. Edge Copilot ships on by default with new Windows 11 installs, putting it in front of every enterprise buyer who upgrades. Treat the surface as material for technical buyer personas today and material for all personas inside 12 months.

Does ranking in Google still help inside Comet and Atlas?

It helps for the tertiary fallback pool, which fires about a third of the time. It does not help for the active-tab or session-history pools, which are responsible for most citations. A page that ranks first in Google but never gets clicked by the buyer is invisible inside the AI browser.

Should I optimize my page differently for the AI browser DOM extractor?

Yes. Move the answer above the fold, render in HTML rather than client-side JavaScript, and keep total time-to-interactive below 2.5 seconds. The same fixes also help organic SEO and AI Overviews, so the effort compounds across all three surfaces.

Which AI browser should I audit first?

Audit Atlas if your buyers skew technical or product. Audit Comet if they skew research and analyst. Audit Edge Copilot if they skew enterprise IT. Run all five if your sales motion crosses categories. The audit takes about four hours per product and the diagnosis transfers across them.

Does AI browser citation work replace my existing AEO program?

It extends it. The same content fixes, schema work, and earned-media program that drive ChatGPT and Claude citations also drive AI browser citations in the tertiary fallback pool. The only net-new work is the DOM extractability layer and the bookmark-pool content investment.

Bottom line

AI browsers are a third citation surface, distinct from chatbots and AI Overviews, and most B2B brands have never tested whether their pages appear inside one. The retrieval model is different in ways that punish brands optimized only for SERP-style ranking: the active tab matters most, the session history matters next, and the web index is a fallback that fires roughly a third of the time.

The fix runs in three layers. Make your pages extractable inside 2.5 seconds. Build content buyers actually return to. Earn the tertiary index citations through the same Wikipedia, Reddit, YouTube, and trade-press program that already drives ChatGPT and Claude citations. Re-audit quarterly. Brands that run this loop will see measurable AI browser citation share inside two quarters. Brands that ignore the surface will keep watching their buyers get synthesized recommendations that exclude them entirely.