How to Show Up in Perplexity Answers (2026)
How we get clients into Perplexity citations in 2026: PerplexityBot and IP allow-lists first, then Promptwatch on the answer. We do not sell a schema package for this.
Perplexity is the engine clients mention after ChatGPT. We still start with their crawler docs, not a brief titled "write like Perplexity." The platform we measure on is Promptwatch. Showing up means a numbered source they control. A name-drop without a link is a mention. We will not call that a citation in a QBR.
Security call
PerplexityBot allowed in robots.txt. IPs from perplexitybot.json. Up to 24 hours. Full UA includes PerplexityBot/1.0. Perplexity says Bot is for search surfacing, not foundation-model training.
Perplexity-User may ignore robots.txt because a user asked. IPs: perplexity-user.json. We explain that before someone "blocks all AI" and wonders why live questions still hit origin. Drafts that must stay private need auth, not Disallow.
WAF rules: user-agent and IP together, as Perplexity wrote for Cloudflare and AWS. Allow when UA contains PerplexityBot or Perplexity-User and the source IP is in the JSON. We schedule JSON refreshes. We do not maintain a handwritten IP wiki. We watch logs after a rule change because they say propagation takes time.
Editorial
A page that originates the comparison or the procedure, claim in visible HTML. We already use Google's quality language with clients. We do not pretend Perplexity signed it. We do not sell llms.txt as their admission ticket. That crawler page does not make one.
22M+ monthly active users is context from the Promptwatch sheet, not a forecast.
Measurement
Paid Promptwatch includes Perplexity. Citation analytics show the URL, including when Reddit or YouTube wins. Referrals appear in visitor analytics through the script or GTM. Explore is ChatGPT-only, so the Perplexity clause in the SOW starts at Essential ($95/mo). Essential has no listed crawler-log allowance. Professional ($245/mo) is the first brand plan with crawler logs, including PerplexityBot, and it also expands the prompt list. Paid checks refresh daily. They are not a pager. We import GSC queries as prompt candidates. We do not treat GSC generative impressions as Perplexity cites. They are different products.
Otterly can sit as a cheap Perplexity mention check from $29/mo. Possible week lag. It is not the client log. Profound Starter is ChatGPT-only. Wrong SKU. Perplexity's 22M+ monthly active users (Promptwatch sheet) is context, not a traffic forecast we put in a deck.
If Bot never 200s, we do not write. If Bot 200s and Reddit is the source, we discuss offsite, not a tenth blog. If an old URL wins, we fix that URL. Frozen prompts from sales, two weeks, cited-host column.
We also wait the documented window after robots.txt (up to 24 hours) before we call the allow-list dead. A same-day "it still 403s" Slack is usually cache or a WAF rule that never loaded the new JSON. We screenshot the Cloudflare or AWS allow (UA contains PerplexityBot or Perplexity-User, IP in the published ranges) into the ticket. We do not paste CIDRs from memory.
Mention without a numbered source stays in a separate column. utm referrals, if they exist, stay in a third. We will not blend those into one "Perplexity score" for a QBR. Content Agents stay closed until Bot 200s. WordPress: write in the CMS they already use. That is how we show up without lying about control.