Perplexity Decider, Teammates and the Amazon Case: AEO Impact

Perplexity's biggest AEO-relevant release in the past 30 days is Decider v1.3, published on 12 November 2026: open-weight models that, by Perplexity's own description, can rerank search candidates, decide whether a query needs web search, and score crawled pages for quality. It also launched Teammates agents on 4 November, and its Amazon Comet case continues. I found no new announcement on its publisher program, shopping or ads. Current as of 13 November 2026.

Key Takeaways

• Decider v1.3 and Decider Flash v1.3 are open-weight under an Apache license and available on Hugging Face, OpenRouter and the Perplexity Decisions API, according to Perplexity's post dated 12 November.
• The post lists reranking, query classification and crawled-page quality scoring as uses, including filtering spam, error pages and anti-bot walls. The numbers (83.99 mean nDCG@10 across 72 retrieval tasks) are Perplexity's own benchmarks, not independent tests.
• Teammates, announced 4 November, are agents for ongoing work, available first to Perplexity Max subscribers on the web, each running in its own environment with separate browser sessions (Perplexity).
• On 4 August 2026 the Ninth Circuit vacated the injunction that had blocked Comet's shopping agent on Amazon, holding that the user, not the agent, "accesses" the site (Jones Day). A hearing on Perplexity's motion to dismiss Amazon's amended complaint is listed for 20 November.
• The Perplexity blog index I read, covering 15 October to 12 November, shows no posts on publishers, shopping, Comet or a search API. Absence of a post is not proof of no change, but I would not claim one.
• The practical AEO action is the same as before: let PerplexityBot reach your pages and check that your CDN is not challenging it.

What Perplexity shipped, by date

Decider: what Perplexity says it does

Perplexity describes Decider as a general-purpose classifier. It takes an input and a set of questions and returns a probability for each possible answer. It does not generate text. The post lists several uses relevant to search:

• Reranking. It scores query and document pairs for relevance and reranks the top 50 candidates from embedding retrieval.
• Search triggering. It classifies queries, for example whether one should trigger web search, and can estimate research effort before search begins.
• Crawl quality. It scores crawled pages for quality and filters out spam, error pages and anti-bot walls.

The benchmark claims, all Perplexity's: 91.6% mean accuracy across 63 query-classification tasks, 6.54 points above its production classifiers; 83.99 mean nDCG@10 across 72 retrieval tasks, 1.81 points above Qwen3-Reranker-8B; and 93.98% agreement with LLM judges on sampled production requests. FourWeekMBA reports that the models are free on OpenRouter until 19 November, citing OpenRouter's CEO, and that its cost figures are the writer's own arithmetic.

The post does not discuss citations. It also does not say, in the version I could read, which of these uses run on every live Perplexity answer. So I would not tell a client that Decider decides who gets cited. What I would say is that Perplexity has published a ranking and page-quality component, and the "anti-bot walls" mention is a reminder that pages behind challenges may be filtered out before ranking. That matches the retrieval picture in how Perplexity retrieval works: getting retrieved and judged relevant comes before anything about your brand.

Teammates and what agents do to your site

Teammates build on Perplexity Computer. According to Perplexity's post, they plan and coordinate long-running projects, can run Computer sessions and other teammates, and monitor work without a new prompt. Access starts with Max subscribers, web first, with iOS "coming soon." The post does not describe web search or citations for Teammates, so any claim about how they cite sources is unsupported. The Tech Buzz, which dates its report 5 November and gives no pricing, does not cover it either.

For site owners the relevant point is traffic type. Perplexity's bot documentation separates two user agents. PerplexityBot surfaces and links sites in Perplexity search results, and Perplexity recommends allowing it in robots.txt. Perplexity-User handles user-initiated fetches and, per the docs, "generally ignores robots.txt rules" because a user requested the visit. Agents acting for users will look like the second category. Perplexity also recommends matching both the user agent string and published IP ranges in a WAF, and warns that Cloudflare Bot Fight Mode can still challenge matching bots. I wrote about what breaks in these visits in what breaks when an AI agent visits your website.

The Amazon case: where it stands

The Ninth Circuit reasoned that Comet's user directs the agent through their own browser, so the user accesses Amazon's computers, not Perplexity. Jones Day notes this was a preliminary-injunction ruling on a limited record, and that a site cannot rely on the CFAA to bar a user-directed agent just because it sent a cease-and-desist letter. Amazon's new allegations are allegations, not findings. For marketers, the takeaway is narrow: sites that want to restrict agents will need technical controls and terms, not an expectation that computer-access law will do it. Background on the commercial side is in agentic commerce reality check.

Publishers, ads and the rest of the legal docket

• Dow Jones v. Perplexity. The tracker lists fact discovery ending 10 December 2026 and a telephone conference on 14 December, and says plaintiffs' motion to compel was granted "in full" on 29 October under a sealed order. The tracker's table header carries an older "as of" date, so treat that entry as partly unverified.
• CNN v. Perplexity. Perplexity's counsel told the court on 23 October that briefing on its motion to dismiss was complete. No hearing date is set.
• Comet Plus. Engadget reported on 25 August 2025 a $5 monthly subscription, 80% of revenue to publishers and a $42.5 million initial pool. I found no 2026 payout totals or subscriber counts from a primary source.
• Ads. Campaign Asia, crediting the Financial Times, reported Perplexity dropping ads, with executives warning ads risk leaving users "suspicious of everything." The page I read gives no date or timeline, and other outlets describe the move differently (a pause versus a wind-down), so I would say only that Perplexity has stepped back from ads.

For the wider legal picture, see AI search lawsuits 2026.

What this means for marketers and founders

• Check PerplexityBot access. Test your key pages with a request using its user agent from outside your network. If your CDN returns a challenge, allow-list its published IP ranges. This is the only Perplexity-specific change I would make this week.
• Keep content retrievable and plain. Decider's description of filtering error pages and anti-bot walls is a reason to avoid interstitials and heavy bot challenges on pages you want cited.
• Do not rebuild strategy around Teammates. Max-only, no stated citation behavior, no pricing for broader rollout. Watch it, do not optimize for it.
• Keep measuring citations separately from mentions. The formats that win on Perplexity are covered in how to get cited by ChatGPT and Perplexity.
• Leave "Perplexity ads" off your plan. There is nothing to buy that I can verify.

Caveats

• I read Perplexity's two posts through a page summarizer, not a raw copy, so check the originals before quoting them.
• The legal dates come from a tracker built on docket entries; the filings themselves were not readable.
• Perplexity's benchmark numbers have no independent replication that I found.

FAQ

What did Perplexity announce this week?

On 12 November 2026 Perplexity published Decider v1.3 and a smaller Flash version as open-weight models under an Apache license. They are available on Hugging Face, OpenRouter and the Perplexity Decisions API.

Does Perplexity Decider change how websites get cited?

Perplexity's post describes reranking, query classification and page-quality scoring, but does not discuss citations. It is a ranking and filtering component, and I have no evidence of how it changes which sources are cited.

What are Perplexity Teammates?

They are agents for ongoing work that coordinate projects, run Computer sessions and monitor tasks without new prompts. They launched on 4 November 2026 for Max subscribers on the web, with iOS planned.

Should I allow PerplexityBot in robots.txt?

Perplexity's documentation recommends allowing it so your site can appear in Perplexity search results. It states that PerplexityBot is not used to crawl content for foundation model training.

Does Perplexity-User follow robots.txt?

Perplexity says it generally ignores robots.txt rules because a user requested the fetch. Access control for it has to happen at the site or CDN level.

Is the Amazon v. Perplexity case over?

No. The Ninth Circuit vacated the injunction on 4 August 2026 and Amazon has since filed an amended complaint. A hearing on Perplexity's motion to dismiss is listed for 20 November, per a docket-based tracker.

Is Perplexity still running ads?

Reports say it has stepped back from advertising. The sources I read differ on whether this is a pause or a wind-down, and none gave me a firm timeline.

Want help with this?

I'm Younus Fardeen, an organic growth, SEO and AEO strategist with 4+ years of marketing experience across edtech and startup teams. If you want help testing whether Perplexity and other AI search bots can reach and cite your pages, see my work and reach me through the contact form at younusfardeen.com.