ChatGPT and Perplexity can answer the same question while citing almost entirely different sites. In one study, Perplexity did not cite 89.1% of the sites ChatGPT cited. Same query, different winners. I spent enough nights mining search term reports to recognize the mistake here: treating two systems as one channel and wondering why half the work goes nowhere.

This list is for owners and marketers who want their business named in AI answers, not another instruction to “do GEO.” ChatGPT’s hidden searches favor a self-contained answer; Perplexity leans on fresh, crawlable pages and corroborating sources. Each entry gives you a page or access change to try, the reason it might work, and a first step. I’ve ordered them by the value of the opportunity, not by how quickly you can tick them off.

1. Publish an honest X vs Y comparison for ChatGPT

When a buyer asks whether to choose you or a named alternative, ChatGPT has a use for a page that answers that question directly. In one HubSpot analysis, comparison content reached a 95% citation rate on ChatGPT, the highest of any format on any engine in that study. X vs Y titles led on ChatGPT; how-to titles led on Perplexity. That does not mean Perplexity refuses comparison pages. It means the same page format is not an equally strong bet on both engines.

The mechanism is answer fit. ChatGPT can find a page that already puts the alternatives, differences, and verdict in one place. Perplexity can retrieve a wider set of candidates and assemble its own account from them. A comparison page gives ChatGPT less stitching to do.

I used to tell clients comparison pages were middle-of-funnel fluff. I was wrong. For a buyer asking us versus them, they are useful inventory. Start with one page that names the alternative honestly, explains three concrete differences in price, fit, or process, and says who should not choose you. Don’t bury the answer beneath a sales pitch. Give the reader a verdict they can inspect, not a disguised ad.

2. Make a named best-for list both engines can quote

A buyer who asks for the best CRM for plumbers or the best garage door company in Dallas is asking for names. Both engines need pages that supply them. Listicles accounted for 21.9% of AI citations in one content-format study, which makes them a sensible cross-engine format rather than a trick for one model.

Why do they travel well? A short block for each named option gives a retrieval system a clear unit to extract. The option, the fit, and the reason sit together. “We serve businesses of every size” gives it nothing comparable to work with.

Start with one best [service] for [industry or city] page. Put three to five named picks on it, including yourself, and give each a distinct fit. If another business is a better fit for a particular buyer, say so. That honesty makes the page useful; pretending every buyer belongs with you defeats the format. Name the options and explain who each is for.

3. Refresh the facts and show when you did it

Perplexity’s retrieval-first approach gives fresh, source-ready pages a reason to compete for its shortlist. One account of its citation process describes 60-plus candidates, 3 to 4 inline citations, and 6 to 9 references per answer. A visible last-updated date and recent statistics are useful signals when those candidates cover a changing topic. In the format analysis that found the comparison effect, last-updated dates performed better than original publication dates.

The date is not the work. If a service page still carries an old price or an outdated process, adding Updated today only makes the mismatch easier to spot. Freshness has to describe the page, not decorate it.

Start with your five most important service pages. Check the facts a buyer could use to make a decision, correct what has changed, and add a human-readable updated line near the top. If a number no longer belongs there, replace or remove it rather than changing the date around it. ChatGPT can still use a clear older answer; Perplexity gives you more reason to keep the page current. Refresh the fact first, then stamp the date.

4. Put one complete answer where ChatGPT can find it

ChatGPT’s hidden searches need a passage that survives being pulled away from the rest of the page. A paragraph that names the service, answers the question, and supplies its key detail can stand alone. A paragraph that begins “As mentioned above” cannot. Perplexity may still combine that passage with other sources, so this is chiefly a way to make your own answer easier for ChatGPT to use.

There is a broader extraction case for structure, too. The content-format analysis reports that lists and tables improve extraction by about 43%, while a summary of the Princeton GEO test reports visibility gains of roughly 25% to 40% from adding statistics, quotations, or citations. Those figures describe different tests. The practical point is narrower: put a specific, supportable answer in a form that can be lifted without its surrounding paragraphs.

Start with your main service page. Near the top, write a roughly 50-word paragraph answering the buyer’s most direct question. Include a relevant number and date if the page has real ones, then put any useful specifications in a table below. Read the paragraph on its own. If it needs a headline, a footnote, and three earlier paragraphs to make sense, rewrite it. Make the first answer complete enough to quote.

5. Get the same fact stated somewhere other than your site

Your own site can establish what you offer. It cannot, by itself, provide an independent view of you. One off-page citation analysis found brands 6.5 times more likely to be cited through a third party than through their own sites and attributed 84% of citations to earned media. In its comparison, identical content received an 8% citation rate on a company blog and 34% through news outlets. That is a reason to care where a fact appears, not a reason to abandon your own pages.

The engines use that material differently. An analysis of 1,765 answers and 13,184 citations found Perplexity drawing on a wider set of sources, including LinkedIn, YouTube, and Reddit, while ChatGPT leaned more on primary pages such as documentation and vendor pricing. The source counts in that analysis are not the same measure as the reference counts in entry three. Both point to the same useful distinction: Perplexity has more room to corroborate what you say elsewhere.

Start by publishing one clear factual page on your domain. Then look for appropriate places where that fact can be stated accurately by someone else: a relevant directory, a partner post, or a video page. Don’t copy a slogan across three profiles and call it consensus. Keep the primary fact on your site; earn independent confirmation off it.

Separate library drawers holding sources for ChatGPT and Perplexity

6. Put usable numbers on the pricing page

A page that says only contact us for pricing leaves a buyer with a question and an AI answer with a blank. A page with a real starting price, what it includes, and when the information was checked gives both of them something concrete to use. The draft research found pages with pricing, ratings, and specifications cited more often than vague equivalents. What the deck calls extractability, I call giving the model something it can repeat without guessing.

ChatGPT’s preference for primary sources makes the vendor pricing page especially important: it can use your stated number rather than a third party’s interpretation. Perplexity may also look for the same detail elsewhere. Neither engine can quote a price you have chosen not to publish.

Start with the smallest number you can state accurately: a starting price, a clearly defined package price, or another real figure that belongs on the page. Say what is included and when you last checked it. If you cannot publish a fixed price, don’t invent one for the sake of a citation. But if you can, stop making people request a call to learn it. A usable number beats “flexible pricing.”

7. Check search-bot access before rewriting another page

Content cannot win a citation from an engine that cannot reach it. The bot names matter here. GPTBot is associated with training; OAI-SearchBot feeds ChatGPT search answers. PerplexityBot serves Perplexity search and is not used for training. A rule intended to stop training can therefore also interfere with search if it blocks every bot with an AI-related name.

Free tool: the AI Crawler Checker shows which AI crawlers your robots.txt lets in and what they receive when they arrive.

This is the mechanical check I wish more people made before commissioning ten new pages. Open robots.txt and inspect the rules for OAI-SearchBot and PerplexityBot. If you want search access while blocking OpenAI training, distinguish OAI-SearchBot from GPTBot; don’t write one sweeping rule and hope it understands your intent. Also check whether a broader rule already blocks the pages you want cited. A training decision and a search-visibility decision are not the same decision.

If you need a fuller permission check, start with the groas guide to fixing access before content. Make the bot check first. A better paragraph will not repair a blocked route to it.

8. Fix the listings and Bing coverage behind your business

ChatGPT Search draws on a candidate set that can include Bing-indexed pages before a live browse and cited answer. That makes Bing visibility worth checking even if you spend your working day watching Google. Perplexity’s wider source mix creates a different reason to care about business listings, reviews, and other outside profiles: they give it places to find the same business details beyond your own testimonial page.

Start with the foundation you control. Claim Bing Places and submit your sitemap in Bing Webmaster Tools. Then review your six most important directory profiles for a consistent business name, category, hours, and service area. This is not a vote-counting exercise where ten listings automatically outweigh one good page. Inconsistent details simply make the answer harder to verify, while absent pages are harder to discover.

A tidy blog cannot compensate for a business that search systems struggle to locate or identify. Check discovery and identity before you blame your copy.

Shopfront beside review cards and a business directory listing

9. Put the city and the job together, then test the split

For a local service query, the words on the page have a job. A page that says garage door repair in Dallas and supplies relevant local details is easier to match to that request than one that says only serving the greater metro area. Perplexity’s retrieval process gives that clear wording a reason to matter. ChatGPT may also find the page through its candidate sources, but that is not a reason to make it guess what the location page covers.

I would start with the top three location pages. Put the city, exact service, and one genuine local proof point in the first 60 words of each. Keep the rest of the page readable; repeating the city in every sentence is not proof of anything. Say plainly what job you do and where you do it. Then test whether the change affects citations rather than assuming it did.

Question: Does clearer local wording help Perplexity cite your pages more often than ChatGPT does?

  1. Set the baseline. Use these nine buyer-prompt types as a starter, then add 11 prompts your buyers might use: three comparisons, three best-for questions, three pricing or process questions, and two local questions. Ask all 20 in ChatGPT Search and Perplexity, logged out, on the same day. Record every citation, its page, and whether your business appears.
  2. Change one thing. Rewrite the local wording on those three pages. Leave the other pages unchanged for two weeks. If the bot check or Bing check reveals a problem, fix it, but log that intervention separately; otherwise you will not know what moved the result.
  3. Run the same prompts again. Keep the wording and conditions as consistent as you can. Compare citations by engine and page, especially on the local prompts. Count appearances, not clicks. This is a small directional check, not proof that one paragraph controls an engine.

My expectation is that Perplexity will respond more clearly to explicit place-and-service wording. The mechanism is straightforward: it has better matching text to retrieve. If only ChatGPT moves, I would look harder at which page it found and whether the answer paragraph did the work. If neither moves, I would check access, Bing coverage, and outside mentions before rewriting the location copy again. Either result changes the next task.

This is the kind of work I do inside groas now. The old agency model charged a percentage of spend for periodic account reviews and called the whole bill strategy; groas runs paid and organic search continuously, with a named strategist accountable for direction, for a flat fee. But I would not start by buying a bigger content plan. I would start with the location pages: they are a contained edit, and the two-engine prompt log tells me where to spend the next week.