November 18, 2026 · ChimpanSEO

What did Perplexity actually release with Decider v1.3?

Perplexity’s Decider v1.3 is a decision model, released on 12 November 2026, that Perplexity says it uses to judge page quality during crawling, to decide whether a query triggers web search at all, and to rerank search results. For SEO and GEO practitioners, the practical takeaway is simple: a page that carries no useful information on its own can be filtered out before any ranking or citation question even arises. That shifts attention from keyword coverage to self-contained, genuinely informative pages.

What did Perplexity actually release with Decider v1.3?

Perplexity released Decider v1.3 and Decider Flash v1.3 as decision models available on Hugging Face under the Apache license, through OpenRouter, and via the Perplexity Decisions API.

According to Perplexity’s research blog, both models ship as open artifacts, which means anyone can download and run them. The company describes three distinct jobs for Decider: judging page quality during crawling, deciding when a user query should trigger web search, and reranking retrieved documents. Perplexity also reports its own evaluation of Decider as a reranker, with a mean nDCG@10 of 83.99 across 72 retrieval tasks. That figure is Perplexity’s own claim, not an independent benchmark.

How does Decider judge page quality during crawling?

Perplexity says Decider judges the quality of pages as it crawls them, flagging low-value pages so they can be filtered out of its index.

The research blog lists the kinds of pages Perplexity considers too low quality to be useful. It is worth reading that list carefully, because it names categories that many sites still publish at scale:

  • Anti-bot walls and CAPTCHAs served to crawlers
  • Error pages
  • Pages that load correctly but contain no helpful information
  • Spam and SEO filler
  • Navigation-only pages

Perplexity also says that in the same crawl pass Decider can identify the domain and intended audience of the content, and whether it contains harmful, illegal, or personally identifiable information. That is a lot of judgment packed into a single crawl step, and it reframes crawling as a quality gate rather than a simple fetch.

Why does Decider matter for SEO and GEO in 2026?

Decider matters for SEO and GEO in 2026 because Perplexity describes several gates a page must pass before it can ever be cited in an answer.

Our reading is that these gates stack in sequence. First, a quality judgment at crawl time. Second, a decision on whether the query triggers web search at all. Third, a relevance score for each query-document pair, scored independently. A page that fails the first gate never reaches the third. For publishers, that means the old habit of publishing thin pages to “cover” a topic carries a different kind of risk than it did under classic ranking systems.

Gate What Perplexity says Decider does
Crawl Judges page quality while crawling and filters low-quality pages from the index
Query routing One head of a multi-head classifier decides whether a query triggers web search
Reranking Scores each query-document pair independently to rank the top 50 retrieved candidates

What are the practical implications for content teams?

Content teams should treat every important page as a standalone answer, because Decider scores each query-document pair independently rather than in the context of the wider site.

That independence is the detail worth internalizing. If a page only makes sense after reading three other pages, it is a weak candidate under a per-pair relevance score. The same logic applies to crawl access: Perplexity lists anti-bot walls, CAPTCHAs and error pages among low-quality pages, so a page that blocks or breaks for crawlers can drop out of consideration entirely. And because not every query triggers web search, some questions may never produce citations regardless of how good the content is. Our reading is that this makes query routing a factor teams cannot fully control, which argues for focusing effort on the pages that do get retrieved.

What should you do today to prepare for Decider?

You should audit your site for pages that exist mainly to navigate or to pad keyword coverage, then either add substance or consolidate them.

  1. Find pages that are mostly navigation, boilerplate or filler, and give them real content or merge them into stronger pages.
  2. Check what Perplexity’s crawler actually receives, including status codes, bot challenges and rendered content, rather than what a browser shows you.
  3. Make each important page self-contained: state the answer, the context and the evidence on the page itself.
  4. Optionally run key pages through the open Decider model as a rough quality check, treating the output as indicative only.

1 Find pages thatare mostly navigation,… 2 Check whatPerplexity’s crawler… 3 Make eachimportant page… 4 Optionally run keypages through the…

Because Decider is openly available under the Apache license, running a similar classifier on your own pages is technically feasible. Whether such a local run matches Perplexity’s production setup is not something the research blog states, so treat any score as a signal, not a verdict.

What should you avoid claiming about Decider?

You should avoid treating Decider as a published ranking factor list, because Perplexity’s post does not set out quality thresholds or weights.

Three traps are easy to fall into. First, do not present the nDCG and accuracy figures as independent results: they are Perplexity’s own evaluations. Second, do not assume Google or ChatGPT apply the same filter, since each engine runs its own pipeline. Third, do not claim that a Decider score predicts citations in Perplexity answers. Our reading is that the model tells you something about page quality as Perplexity defines it, and nothing more specific than that.

Frequently Asked Questions

Is Decider v1.3 free to use?

Perplexity says both Decider v1.3 and Decider Flash v1.3 are available on Hugging Face under the Apache license, through OpenRouter, and via the Perplexity Decisions API. The Apache license is a permissive open source license, so you can inspect and run the models yourself.

Does Decider replace traditional ranking signals?

Perplexity describes Decider as a decision model used for page quality judgment, query routing and reranking. The research blog does not present it as a replacement for other ranking signals, so treat it as one component in a larger retrieval pipeline.

Can a good page still be left out of Perplexity answers?

Yes. Perplexity says one head of its query classifier decides whether a query triggers web search at all. If a query does not trigger search, no page gets cited for it, however strong the content may be.

How do I check whether my pages pass a quality filter?

Start with crawl access. Confirm that crawlers receive a normal status code and rendered content, with no bot challenge or error page. Then review each important page and ask whether it answers its target question on its own, without needing other pages for context.

Our reading

We think Decider signals a shift that publishers should take seriously: quality judgment is moving earlier in the pipeline, to the crawl itself. If a page is mostly navigation, filler or boilerplate, it may be filtered before relevance is ever scored, which makes self-contained pages the safer bet. We would not over-read the numbers Perplexity published, since they are the company’s own evaluations. But we would expect teams that audit crawl access and consolidate thin pages to be better positioned as these gates tighten through 2026.

Sources

Related reading

This article was written and published with ChimpanSEO

Write SEO/AEO articles and publish them to WordPress: it usually takes 1-2 minutes. Try it free, no card required.

Related articles