Guides / Perplexity citations
Why Perplexity cites your competitor and not you, and how a UK content lead fixes it
By Olly, founder of AEOscore, AEO consultant and full-stack developer, 20+ years in search13 min read
Why does Perplexity cite our competitor and not us?
On this page
Original data
Perplexity names who it has read: our numbers
Across 1,297 Perplexity answers to 107 buyer questions for 3 UK businesses we track, Perplexity listed about 17 sources per answer. In the 694 answers that named a rival and not the business we track, the rival’s own website was one of those sources 636 times: 92%.
So when Perplexity names your competitor, it has almost always read your competitor’s page. That fits how Perplexity describes its own engine: documents are split into short passages, and each passage is retrieved and ranked against the question on its own [2]. A business with no passage that answers the question gives Perplexity nothing to name it from, however good its homepage is.
Most of what Perplexity reads is neither your site nor a tracked rival’s. Over 80% of the sources were third-party pages, and about one in five were “best”, “top” and “vs” lists. Those lists are where a rival picks up the mentions that keep it in the answer week after week.
Each Perplexity answer listed between 9 to 38 sources, a median of 20. Through its API the count was up to 20 a time, now and then more, until late September 2026, then 9 or 10 for every answer in the 30 days to 2 November 2026. We changed nothing on our side, so fewer pages now make each answer’s list.
Every Perplexity answer with text, from AEOscore's scans of three UK businesses in different sectors (one of them AEOscore itself) between 3 August and 2 November 2026, asked through Perplexity's API in UK buyer phrasing. Two of the three were last scanned in mid-October. A rival counts as named when the answer names a business on that site's tracked competitor list, and its website counts as cited when a source URL sits on that rival's tracked domain. Rivals we do not track, and second domains of tracked rivals, count as third-party pages, so both rival figures are more likely to be under than over. Sources were sorted by domain against fixed lists of review sites and of forum, video and social sites; the remaining third-party URLs were split by a path pattern (best, top, alternatives, vs, review, comparison) into list pages and everything else.
Every engine
Does every AI engine cite the same sources?
No: in the same scans, each engine read a different mix of sources and named different businesses, so a Perplexity citation tells you little about ChatGPT.
| Engine | Answers | Median sources | No sources | Your own site | Rivals' sites | Lists and comparisons | Review sites | Forums, video, social | Other third-party |
|---|---|---|---|---|---|---|---|---|---|
| Perplexity | 1,297 | 20 | 0.0% | 3.5% | 13.1% | 20.6% | 2.2% | 4.4% | 56.2% |
| Gemini | 1,170 | 14 | 6.4% | 6.4% | 11.9% | 18.3% | 1.7% | 7.2% | 54.5% |
| Claude | 324 | 14 | 18.2% | 3.8% | 16.1% | 19.1% | 4.2% | 0.8% | 56.0% |
| Google AI Overviews | 1,258 | 10 | 7.7% | 4.3% | 10.4% | 17.8% | 0.9% | 11.1% | 55.6% |
| Google AI Mode | 1,289 | 8 | 4.4% | 4.0% | 11.3% | 15.1% | 0.9% | 10.4% | 58.2% |
| ChatGPT | 1,297 | 0 | 58.3% | 10.6% | 13.1% | 9.2% | 1.3% | 9.9% | 55.8% |
3 August to 2 November 2026, the same three UK businesses. Answers collected through each engine's API or search-surface data, so a consumer app can show different sources. Failed answers are left out. Claude has fewer answers because it was not scanned for part of the window. Percentages are shares of all sources each engine cited. "Your own site" is the tracked business's domain; "rivals' sites" are the domains on its tracked competitor list.
The engines also disagree on who to name. Of the 3,247 times Perplexity named a business we track, ChatGPT named the same business in its answer to the same question in the same scan 27% of the time; Gemini 42.2% and Google AI Overviews 32.4%. ChatGPT listed no sources at all in 58.3% of its answers, while every Perplexity answer listed at least nine.
Perplexity is also a small engine in the UK next to Google, where roughly 30% of keyword searches showed an AI Overview (Ofcom, December 2025; more in our UK AI search statistics). Fix the Perplexity gap with the steps below, then check each engine on its own.
Diagnose
Seven reasons Perplexity reads them and not you
Every reason fails one of three tests, fetch, lift or confirm, and you work through them in that order, because a failed earlier test stops the later ones from mattering.
Test 1 of 3
Can it fetch you?
Perplexity cites pages its crawler can reach and read as text; a blocked page, or one whose text only loads by script, rarely gets picked.
1. Perplexity cannot read your page
PerplexityBot is the crawler that feeds Perplexity's search citations, and Perplexity says it obeys robots.txt [2]. A disallow line, a firewall rule or a bot-protection setting keeps your page out of the index while your rival's page stays in. Content that only appears after JavaScript runs is a second trap: in one 2026 test of the live fetch Perplexity makes when a user shares a link, it reported the page's raw HTML and never saw what the script added [11].
Check: Run our robots.txt and live access test for PerplexityBot, then view the page source (not the rendered page) and confirm the answer text is there.
Fix: Allow PerplexityBot in robots.txt and at the firewall, and serve the key copy in the HTML. Perplexity says robots.txt changes can take up to 24 hours to apply [1].
“Most of the access problems I find were never a decision. Someone switched on bot protection years ago, and nobody in marketing gets an error when Perplexity is turned away.”
Test 2 of 3
Can it lift an answer?
It builds answers from short passages, so the page needs a clean answer near the top.
2. You have no page for the question
Your rival usually has a page built around the buyer's question: a comparison, a pricing page, a page for one type of customer. You have a homepage and a services page that mention the topic in passing. Perplexity retrieves passages that match the question, and a page that never asks it rarely holds one.
Check: Search your own site for the question in the buyer's words. If the closest match is your homepage, that is the gap.
Fix: Publish one page per losing question, titled in the buyer's words, with the answer in the first two sentences.
3. Your answer only makes sense in context
Perplexity splits documents into "self-contained spans, each of which can be individually retrieved and ranked at query time" [2]. A section that relies on the paragraph above it, or names the business only in the page header, loses to a rival's section that answers on its own.
Check: Copy one section of your page into a blank document. If a stranger could not tell who you are, what you offer and what it costs from that section alone, it will not stand up as a passage.
Fix: Use the question as the heading, answer it in the first two sentences, name the business in full, and keep the figures in the same paragraph as the claim.
4. Their page carries more evidence
A widely cited study, the GEO paper from Princeton and IIT Delhi, found that adding citations, quotations and statistics raised a source's visibility in generative answers by 30 to 40% on the paper's main word-count measure [4]. In its Perplexity test, on that same measure, adding quotations did best at 22% over the baseline, and keyword stuffing did 10% worse. The paper dates from 2023 and 2024 and tested supplied pages rather than live search, so treat the figures as indicative.
Check: Count the prices, named products, dates, figures and named sources in the rival's cited page and in yours.
Fix: Add the specifics a buyer would quote: price ranges in pounds, named integrations, delivery times, dated figures with their source, and a short quote from a named customer or expert.
5. Their page looks fresher
Perplexity filters out "clearly non-responsive or stale content" before ranking [2], and Ahrefs found it orders its in-text references from newest to oldest [6]. Age alone is not the test: in the same Ahrefs study, the pages Perplexity cited were older on average than those of every other assistant except Google's AI Overviews, so an older page that is plainly kept up to date can still win.
Check: Look for a visible last-updated date on both pages, and for figures that carry a year.
Fix: Update the page for real, show the date, and keep a short dated note of what changed.
“Unblocking the crawler only makes you eligible. Perplexity still has to find a better answer on your page than on your rival's, and that's a writing job.”
Test 3 of 3
Can it confirm you?
Over 80% of what it cites in our scans is other people's pages, so those pages need to name you and describe you the same way.
6. The pages it reads list them, not you
Most of what Perplexity cites is neither your site nor a tracked rival's: in our data, over 80% of sources were third-party pages, and about one in five were lists and comparisons. Profound measured Reddit at 6.6% of all Perplexity citations [8]. Search rankings overlap with it: Ahrefs found 28.6% of the URLs Perplexity cited also ranked in Google's top 10 [7].
Check: List the third-party pages Perplexity cited for your question and note which ones name your rival and not you.
Fix: Email the authors of those lists with your facts and a reason to include you, answer the Reddit threads where you have something useful to add, and keep your review profiles current.
7. It is not sure you are one business
If your website, LinkedIn, Companies House record, directories and review profiles give different names, addresses or descriptions, the sources Perplexity reads do not agree about you, and a rival that tells one consistent story everywhere gives it less to reconcile.
Check: Ask Perplexity what your business does and read the description it gives. Then compare your name, address and one-line description across your site, LinkedIn, Companies House and your three biggest directory listings.
Fix: Pick one name, one address and one description, publish them on a plain facts page, and correct every listing that differs.
Save your time
What will not fix it
- Adding schema markup. There is no evidence yet that it moves citations. An Ahrefs test of 1,885 pages that were already cited found no meaningful increase in AI Overviews, AI Mode or ChatGPT after schema was added; Perplexity was not tested, and the test could not say whether schema helps pages that are not yet cited [9]. Keep structured data tidy, and put every fact you want quoted in visible text.
- Keyword stuffing. In the GEO paper's Perplexity test, stuffing the page with the query's keywords did 10% worse than leaving it alone on its main measure [4].
- An llms.txt file. We treat llms.txt as stationery: tidy to have, and no reason for Perplexity to choose your page over a rival's.
AEOscore Citation Gap Audit
How does a UK content lead diagnose the gap in one afternoon?
Run the AEOscore Citation Gap Audit: five steps, from 25 real buyer questions to a list of what the rival’s cited pages have that yours lack.
- 1
Write down the 25 questions a buyer types
Use the words UK buyers type, not keywords: pounds, VAT, UK place names, "solicitor" not "attorney". Freeze the list so later checks compare like with like.
What good looks like: 25 full questions a prospect could ask, unchanged from one check to the next.
- 2
Ask each one in Perplexity, logged out, and record every URL it cites
Record the exact source URLs as well as the businesses named, and ask twice: in our own scans the list of brands named changed between days in most question and engine combinations.
What good looks like: One row per cited URL, not just a list of brand names.
- 3
Sort the cited URLs by type
Rival's own site, third-party article or list, directory or review site, forum or video.
What good looks like: You can see whether the rival wins on its own pages or on other people's. In our data, over 80% of what Perplexity cited was third-party.
- 4
Open each rival page beside your nearest page
Read the first 150 words of both, and note what theirs answers that yours does not: the question as a heading, a price, a named product, a date, a quote.
What good looks like: One line per page saying what yours is missing.
- 5
Check your own access
Look at robots.txt and the firewall for PerplexityBot, then view the page source and find the answer text before any JavaScript runs.
What good looks like: PerplexityBot allowed, no bot challenge, and the answer in the raw HTML.
Step two needs a fixed list and patience: answers vary from one run to the next, which is why we measure share of voice over a frozen panel rather than trusting one answer. The short version is read what AI says about your rivals, and the way we measure is on our methodology page.
Template
The log to keep, with a worked example
One row per losing question. The fifth column is where the work comes from.
| Question | Rivals named | Rival page cited | Your closest page | What their page has that yours lacks | Fix | Shipped | Re-asked |
|---|---|---|---|---|---|---|---|
| Which payroll software suits a UK accountancy practice with 20 staff? | Ledgerly and PayPilot | ledgerly.co.uk/payroll-for-accountancy-practices, plus two "best payroll software UK" lists | The Brightwater homepage | Ledgerly's page asks the question in its title, gives a price per payslip in the first paragraph, has a dated comparison table and quotes a named practice manager. The homepage has none of these. | A new page, /payroll-for-accountancy-practices, answering in the first 50 words with the price, the named integrations and an updated date; an email to both list authors with the facts. | Week 2 | Re-asked three weeks later, on the same frozen question. |
A fictional example. Brightwater Payroll and its rivals are made up.
Fix
What does the fix look like, and who does it?
Anyone can run the diagnosis, but someone has to rewrite the pages, change robots.txt and the firewall, and chase the third-party lists, in this order.
- 1
Fix access first
It is a yes or no, and quick: until the crawler can read the page, no content change counts. Perplexity says robots.txt changes can take up to 24 hours to apply.
- 2
Rewrite the first 150 words of your three most important pages
Then build one page per losing question: answer first, the business named in full, the evidence in the same paragraph, a visible updated date.
- 3
Get onto two of the third-party pages it cited
Pick the lists and threads that name your rival for that exact question, and give each author a reason to add you.
- 4
Re-ask after two to four weeks
Perplexity can shift within days, other engines over weeks. Re-ask the questions you froze in step one, and give a fix 4 to 12 weeks before you judge it.
AEOscore does both halves. It asks your buyer questions on all six engines every week (daily on Monitor + Strategist), Perplexity included, stores every source each answer cites, and writes fix plans for the questions you are losing. On Monitor (£99 a month) your team makes the changes and we show whether the answers moved; on Monitor + Strategist (£1,495 a month) a named in-house expert spends a strategist and developer day with you every month and makes them for you. The £395 one-off AI Visibility Audit is credited against your first month.
See what gets fixed, monitoring vs fixing and pricing, or start with the free AEO score to see which questions Perplexity gives to your rivals.
“A dashboard tells you Perplexity picked your rival again. Someone still has to rewrite the page and email the list authors, and that's what moves the answer.”
FAQ
Perplexity citations, asked directly
- Is this a content problem or a technical problem?
- Usually both. Access and rendering decide whether Perplexity can cite you at all; the answer near the top of the page, and the other sources that back it up, decide whether it picks you once it can. Check access first, because it is quick to confirm and quick to fix.
- Does ranking first in Google mean Perplexity will cite us?
- No. Perplexity runs its own index, which it says covers hundreds of billions of webpages. Ahrefs found 28.6% of the URLs Perplexity cited also ranked in Google's top 10 for the same query, so a page that ranks well has a head start, but most citations come from elsewhere.
- How do I find out which pages Perplexity is citing instead of ours?
- Ask your 25 buyer questions in Perplexity, logged out, open every numbered source, and record the exact URL, the type of page and the sentence Perplexity seems to have taken from it. Ask each question twice, because the sources change from one run to the next.
- What should a UK content lead change first?
- Fix crawler and rendering access first, because it is a yes or no and quick to fix. Then rewrite the first 150 words of your three most commercially important pages so each one answers its question outright, with the price, product or date in the same paragraph.
- Does blocking PerplexityBot stop Perplexity citing my site?
- Not reliably. PerplexityBot feeds Perplexity's search results and follows robots.txt, so blocking it keeps your pages out of its index. But Perplexity-User, which fetches a page when someone asks about it, generally ignores robots.txt, and Columbia Journalism Review found Perplexity still identified excerpts from a publisher that had disallowed its crawlers. If you want to be cited, allow PerplexityBot.
- Is schema markup the answer?
- Not on the evidence so far. Valid structured data is housekeeping that keeps a page machine-readable. An Ahrefs test of 1,885 already-cited pages found no meaningful citation increase in AI Overviews, AI Mode or ChatGPT after schema was added; Perplexity was not tested. Put every fact you want quoted in visible text.
- How long before Perplexity starts citing us?
- Access fixes can show within two to four weeks; content and third-party changes usually take 4 to 12 weeks, and neither is guaranteed. On our own site, Perplexity first named us five days after one fix in August 2026, and named us in 65 of 168 answers from mid-October, up from 2 of 40 in August.
- Does this apply to ChatGPT and Gemini too?
- Only partly. Of the businesses Perplexity named in our scans, ChatGPT named 27% in its answer to the same question, and Gemini 42%. ChatGPT listed no sources at all in 58% of its answers, so check each engine separately.
- How many UK buyers use Perplexity?
- Far fewer than use ChatGPT. The Reuters Institute found 6% of UK adults had heard of Perplexity in 2025, against 73% for ChatGPT, and across six countries 1% used it weekly against 22% for ChatGPT. Ofcom reported Perplexity's UK web visits doubled in the year to August 2025. The fixes on this page are largely the ones that get you cited in ChatGPT search and Google's AI answers too.
Sources
Sources
- Perplexity, Perplexity crawlers. Accessed 2 November 2026. PerplexityBot is "designed to surface and link websites in search results on Perplexity" and is not used for foundation models; Perplexity-User "generally ignores robots.txt rules"; robots.txt changes may take up to 24 hours.
- Perplexity Research, Architecting and evaluating an AI-first search API. Accessed 2 November 2026. Published 25 September 2025. Documents are split into self-contained spans ranked at query time; hybrid keyword and semantic retrieval with a reranker; first-cut filters remove stale content; PerplexityBot complies with robots.txt.
- Perplexity, Introducing the Perplexity Search API. Accessed 2 November 2026. Published 25 September 2025. An index "covering hundreds of billions of webpages" processing "tens of thousands of index update requests" each second. Perplexity's own description, not independently audited.
- Aggarwal et al., Princeton University and IIT Delhi, GEO: Generative Engine Optimization (KDD 2024). Accessed 2 November 2026. First published 16 November 2023, revised 28 June 2024. The Perplexity test used 200 queries with source pages supplied as uploads, not live search, and measures how much of the answer a source earns rather than whether it is cited at all.
- Seer Interactive, Study: AI brand visibility and content recency. Accessed 2 November 2026. Published 25 June 2025, 5,000+ cited URLs. "About 50% of Perplexity's citations are from 2025 alone."
- Ahrefs, New study: AI assistants prefer to cite "fresher" content (17 million citations analysed). Accessed 2 November 2026. Published 28 July 2025, about 17 million cited URLs. Perplexity "seems to order citations from newer to older".
- Ahrefs, Only 12% of AI cited URLs rank in Google's top 10 for the original prompt. Accessed 2 November 2026. Published 11 August 2025, 15,000 long-tail prompts. 28.6% of the URLs Perplexity cited ranked in Google's top 10 for the same query.
- Profound, AI platform citation patterns. Accessed 2 November 2026. Published June 2025, updated August 2025; citations from August 2024 to June 2025. Reddit was 6.6% of all Perplexity citations, YouTube 2.0%. A vendor study.
- Search Engine Journal, Schema markup didn't move AI citations in Ahrefs test. Accessed 2 November 2026. Published 11 May 2026, reporting an Ahrefs test of 1,885 pages: "No platform showed a meaningful citation increase after schema was added." Perplexity was not among the platforms tested.
- Columbia Journalism Review, Tow Center, AI search has a citation problem. Accessed 2 November 2026. Published 6 March 2025, 1,600 queries. Perplexity identified all ten excerpts from a publisher that had disallowed its crawlers.
- Search Engine World, Do AI assistants actually render your JavaScript when grounding? We put it to the test. Accessed 2 November 2026. Published 9 June 2026, updated 2 July 2026. Perplexity reported the page's raw HTML and did not run its script. A single test.
- Ofcom, Online Nation 2025. Accessed 2 November 2026. Published 10 December 2025, Similarweb data. Perplexity's UK web visits grew 100% in the year to August 2025, from a much smaller base than ChatGPT's.
- Reuters Institute for the Study of Journalism, Generative AI and News Report 2025. Accessed 2 November 2026. Published October 2025, YouGov survey of about 2,000 adults per country. 6% of UK adults had heard of Perplexity, against 73% for ChatGPT.
£395 one-off AI Visibility Audit, credited against your first month, £99 a month Monitor, £1,495 a month Monitor + Strategist. Prices exclude VAT.
See which questions Perplexity gives to your rivals
The free AEO score asks your buyer questions on all six engines and shows who gets named instead of you, human-checked within 1 to 2 working days. Domain and email, no card.
- ChatGPT
- Claude
- Gemini
- Perplexity
- Google AI Overviews
- Google AI Mode