Perplexity is an AI answer engine that searches the web in real time and writes responses with numbered citations. The most common mistakes brands make with it are blocking its crawler, publishing pages without direct answers, ignoring third-party sources it trusts and measuring results with single screenshots. Each has a practical fix that also helps traditional search.
How does Perplexity choose its sources?
Perplexity runs searches for each question, reads a set of pages and composes an answer that cites them. It uses its own crawler, PerplexityBot, and also draws on other search indexes. Users can switch focus modes, follow up, and in paid tiers choose among several underlying models.
Because answers are built from retrieved pages, the factors that matter are familiar: the page must be reachable, relevant to the exact question and clear enough to quote. Perplexity does not publish a ranking formula, so anyone claiming a guaranteed method is guessing.
Which mistakes do agencies see most often?
Five mistakes come up again and again in Perplexity audits. Most are cheap to fix once someone looks for them.
Mistake 1: blocking the crawler without meaning to
Many sites added broad AI-bot blocks to robots.txt or through a CDN setting, often to stop model training. Some of those rules also block PerplexityBot, which removes the site from Perplexity’s own crawl.
The fix is to decide per bot. Perplexity documents its user agents. Check your robots.txt, and in Cloudflare or a similar service, review the bot management and “AI crawler” settings. Allow the agents you want in search answers and block the ones you do not.
Mistake 2: writing pages that never answer the question
Perplexity quotes sentences. A page that opens with three paragraphs of brand story gives it nothing usable. A page that opens with “Our onboarding takes two to three weeks and includes a kickoff call, data import and staff training” gives it a clean citation.
- Put the answer in the first two sentences under each heading.
- Use headings phrased as the questions people ask.
- Include specifics: numbers you can stand behind, steps, conditions.
- Keep one topic per page so retrieval matches cleanly.
Mistake 3: ignoring the sources Perplexity already cites
For many queries Perplexity leans on review sites, forums, Wikipedia, news outlets and comparison articles. If your brand is absent from those, your own site carries all the weight.
Run your target questions and list the domains cited. If Reddit threads, G2 reviews or industry publications appear repeatedly, plan genuine presence there: answer questions in relevant communities under your real name, keep review profiles current, and pitch data or expert commentary to publications. Do not post fake reviews or astroturf threads; it violates platform rules and tends to be noticed.
Mistake 4: treating one answer as a ranking
Perplexity answers change with phrasing, location, model and time. A screenshot showing your brand at citation one tells you little. Measure across a fixed set of questions, run regularly, and record whether you are cited, where, and which of your URLs is used.
Mistake 5: forgetting freshness
Perplexity often prefers recent pages for topics that change. A pricing guide last updated two years ago may lose to a competitor’s current one. Show a visible last-updated date, actually update the content when facts change, and keep dates in structured data consistent with the page.
How Alien Road applies Perplexity optimisation
In a typical engagement, discovery is a question set of thirty to fifty prompts built from sales calls, Search Console queries and competitor comparisons. We run them in Perplexity and log citations, cited domains and answer accuracy about the client.
Setup starts with technical access: robots.txt, CDN bot rules, server logs to confirm PerplexityBot visits. Then content work on the pages that should be cited, plus a plan for third-party presence based on the domains that keep appearing.
Measurement repeats the question set monthly. In GA4 we track referral sessions from perplexity.ai and their conversions. Reporting shows citation share for the set, the URLs cited, referral traffic and a list of inaccurate statements about the client that need correcting at the source.
How do you correct wrong information about your brand?
Perplexity repeats what it finds. If an answer states an old price or wrong address, find the cited page. If it is yours, fix it. If it is a third-party directory, request a correction from that site. Answers usually improve once the sources change, though not instantly.
A short checklist for accuracy
- Company facts identical across your site, LinkedIn, Crunchbase and directories.
- Old pages with outdated details redirected or updated.
- A clear “About” page with founding details, location and leadership.
- Press releases archived with dates so old news is not read as current.
Which technical checks should you run every quarter?
Beyond robots.txt, confirm that key pages return a 200 status, load their main text without client-side rendering delays and carry a self-referencing canonical. Check server logs for PerplexityBot requests to the pages you care about. If they are never requested, investigate internal linking and sitemap coverage.
Also review your structured data. Organization, Product and FAQ markup do not guarantee citations, but they make facts such as your name, address and offer explicit and consistent. Validate them with Google’s Rich Results Test and Schema.org’s validator after any template change.
Keep a simple log of each check with the date and the result. When citation rates change later, the log shows whether a technical problem was present at the time, which saves hours of guessing during monthly reviews.
One last check is speed. Pages that respond slowly or time out under crawler load may be skipped. Review response times in your server logs for bot traffic, not only for human visitors.
Where to start this week
Check robots.txt and your CDN settings for PerplexityBot. Then write down ten questions a buyer asks before choosing a provider in your category, run them in Perplexity and note every cited domain. Fix the one page of yours that should answer the most common question, and plan one genuine action on the third-party site that appears most often.
Keep reading: GEO – AIO – AEO · GEO