How I Got Cited in 34 Perplexity Answers in 30 Days

In April 2026, I ran a controlled experiment: take five existing articles, apply a specific set of GEO changes, and track how many times Perplexity cited them over the following 30 days. The before state: 3 citations total across all five pages. The after state: 34 citations.

This is a breakdown of exactly what changed, in the order I made the changes, with screenshots of the citation appearances and an honest accounting of what didn't move the needle.

The Starting Point

The five pages going into this experiment were all articles I'd published between 6 and 18 months prior. They had decent Google traffic — averaging around 400 organic visits per month each — but almost no AI citation presence. I was checking them manually in Perplexity once a week and seeing my domain show up maybe once every two weeks across all five.

My baseline across 30 days before making any changes:

I tracked citations by running 12 fixed target queries in Perplexity every Monday and Thursday throughout the experiment, logging which sources appeared in the answer. Each query was chosen because it directly matched the topic of one of the five pages.

How the Test Was Set Up (Methodology)

This was structured as a controlled before/after experiment rather than a loose "I changed some things and citations went up" story. Three details make the results reproducible:

• Fixed query set. The same 12 queries were checked on the same two days each week for the full 30 days — a 60-day window total when you count the 30-day baseline before any changes. Holding the queries constant is what lets the before and after numbers be compared at all.
• One change category per week. Each week introduced a single class of change (dates + direct answers, then FAQ schema, then entity coverage, then outbound links). Staggering them this way is what makes it possible to attribute lift to a specific change instead of guessing — closer to an A/B test than a redesign where everything moves at once.
• Logged every appearance. For each query-day I recorded which domains were cited and in what position within the answer, in a simple spreadsheet (query, date, cited domains, my page yes/no). That log is the raw data behind every number in this article.

The honest caveat: with a five-page sample and manual checking, this is a directional field experiment, not a lab-grade study. The point isn't statistical certainty — it's a repeatable process you can run on your own pages and watch the same signals move.

The Changes I Made (In Order)

Week 1: Dates and Direct Answers

I started with the two fastest changes: adding explicit publish dates to all five pages, and rewriting the opening paragraph of each to lead with a direct, definitional answer to the target query.

Dates: Three of the five pages had no visible publish date at all. I added them to the article header and to the Article JSON-LD schema. The Perplexity crawler re-indexed all five within 72 hours (I confirmed this using Perplexity's site: operator).

Direct answers: Here's an example of the before/after on Page B:

Before: "With the rise of AI-powered search engines, understanding the relationship between traditional SEO and newer optimization approaches has become increasingly important for content creators and marketers alike..."

After: "GEO (Generative Engine Optimization) differs from SEO in one fundamental way: SEO optimizes for ranking in a list of links, while GEO optimizes for being cited as a source inside an AI-generated answer. Both target search engines, but they require different content signals to succeed."

By end of week 1, citations across all five pages had moved from 3 to 9. Page B alone jumped from 1 to 5 — the direct answer rewrite did most of the work.

Week 2: FAQ Sections and Schema

Week 2 was the highest-leverage week of the experiment. I added a FAQPage schema section to all five pages — targeting the exact "People Also Ask" questions that appeared in Google for each topic.

Each FAQ section had 4–6 questions with concise, self-contained answers (2–4 sentences each). I also validated the FAQPage JSON-LD using the Schema.org Validator to make sure there were no markup errors.

The citation jump in week 2 was dramatic: from 9 to 21 total. Pages C and D — which had zero citations at baseline — both started appearing regularly. Perplexity was pulling FAQ answers almost verbatim in several cases, which I confirmed by comparing the Perplexity answer text to my <dd> elements.

Week 3: Entity Coverage Audit

I used Surfer SEO to run an entity audit on each page, comparing my coverage against the top 10 pages Perplexity was currently citing for my target queries. On average, I was missing 8–12 entities per page that the cited pages covered.

I added the missing entities as natural mentions within existing sections — not as new sections, just woven into the prose where they logically fit. For Page C (ChatGPT ranking factors), the missing entities included: "retrieval-augmented generation", "training data cutoff", "browsing plugin", and "semantic similarity scoring."

By end of week 3: 29 total citations. The entity changes produced a smaller lift than the FAQ additions, but they were particularly effective for ChatGPT citations, which I had been tracking separately.

Week 4: Outbound Citation Links

The last week's change was adding outbound links to authoritative primary sources — specifically: research papers, official documentation from OpenAI and Perplexity, and established publications like Search Engine Journal and Wired.

I targeted 3–5 outbound links per page, placed immediately after specific factual claims. The hypothesis was that Perplexity's retrieval model weights pages that themselves cite credible sources — essentially rewarding pages that behave like well-researched articles rather than thin opinion pieces.

End of month 1: 34 total citations, up from 3. Page E — the weakest page going in — ended with 4 citations, which felt like the biggest surprise of the experiment.

Results Summary

What Didn't Work

Honest accounting of the things I tried that produced no measurable citation lift:

• Adding more internal links — I added 15–20 internal links across the five pages based on advice I'd seen elsewhere. Citation count didn't move during the week I made this change.
• Updating meta descriptions — I rewrote all five meta descriptions to be more keyword-rich. No detectable effect on Perplexity citations (makes sense — Perplexity reads body content, not meta tags).
• Adding more images — I added alt-text-rich images to pages C and D. No measurable citation change. AI text retrieval is indifferent to images.
• Increasing word count — I expanded Page A from ~1,400 to ~2,200 words by adding a new section. The citation lift that week was more likely due to the FAQ changes I made at the same time; I can't isolate word count as a driver.

How to Run Your Own Perplexity Citation Gap Analysis

Before touching a single page, I ran a citation gap analysis — and it's the step that decided which five pages to rework first. A Perplexity citation gap analysis is simple: list the queries that matter to your niche, check which sources Perplexity currently cites for each, and flag the queries where competitors appear and you don't. Those gaps are your highest-leverage targets, because Perplexity is already answering them with citations — you just aren't one of them yet.

Here's the exact routine I used:

1. List 10–15 target queries — the real, natural-language questions your pages should answer. Full questions, not keywords.
2. Run each in Perplexity and record every domain cited in the answer, in order. Note whether you appear and, if so, in what position.
3. Mark the gaps — the queries where a competitor is cited and you are not. Sort them by how close the topic is to a page you already have; those are the fastest wins.
4. Prioritize by gap × relevance. A query you're absent from, on a topic you already have a page for, is worth more than one where you'd need to write from scratch. That's how the five pages above got chosen.

The same gap-analysis logic applies across engines. For the ChatGPT side of this — including why some answers cite sources and others don't — see how ChatGPT cites sources, and for the step-by-step playbook, how to get cited in ChatGPT.

How to Replicate This

The experiment is repeatable. Here's the exact sequence, with the time each step took me:

1. Audit your existing pages for visible dates (30 min total). Add publish date and last-updated date to every article header and Article JSON-LD.
2. Rewrite the opening paragraph of each page to lead with a direct definitional answer to the target query (1–2 hours per page).
3. Add a FAQPage schema section using 4–6 PAA questions as H2-level <dt> elements with concise <dd> answers. Validate with Schema.org Validator (2–3 hours per page).
4. Run an entity audit using Surfer or Clearscope. Add missing entities as natural mentions in existing sections (1–2 hours per page).
5. Add 3–5 outbound links to primary sources for factual claims (30 min per page).
6. Set up tracking and a citation gap analysis: define 10–15 target queries, check them in Perplexity twice per week, and log which sources appear. The queries where a competitor is cited and you aren't are your citation gaps — that gap analysis is what tells you which pages to prioritize next.

Total time investment: approximately 5–8 hours per page. The changes I made to these five pages took about 30 hours spread across the month.

For the full breakdown of which GEO tools I used during this experiment, see Best GEO Tools for 2026. For the complete list of signals I now apply to every new article, see The GEO Content Checklist. And for platform-specific playbooks, see how to optimize for Google AI Overviews and get cited in ChatGPT.