What Perplexity rewards differently to ChatGPT for maritime queries

Perplexity and ChatGPT often produce different answers to the same maritime query. Sometimes the brands cited overlap; often they do not. Understanding why points directly at how to influence each.

We run quarterly audits across both, using the same prompt set, and the divergence has stayed consistent enough to be useful. The two engines reward different things, and a brand that is invisible in one is often visible in the other.

What Perplexity rewards

Live retrieval freshness

Perplexity’s default mode is retrieval-heavy and time-sensitive. A maritime article published last week is more likely to be cited than one published two years ago, even when the older article is more authoritative. This favours brands that publish consistently and reward freshness over backlog accumulation.

Cited URL specificity

Perplexity nearly always cites specific URLs, not just brand names. The page must be retrievable, fast-loading and structurally clean. A homepage with a long-form copy block is less useful to Perplexity than a focused service page that answers the specific query.

Direct quotability

Perplexity’s UI shows the user the source passage, often verbatim or near-verbatim. Pages with crisp claim-and-evidence sentences get quoted directly; pages with vague marketing prose get paraphrased and de-emphasised in the citation hierarchy.

Bing-class indexing signals

Perplexity for many query types runs on Bing as the underlying retrieval. Bing’s quirks, particularly its slightly different attitude to schema and to industry directories, leak through. Maritime sites with strong Bing indexing tend to do well in Perplexity.

What ChatGPT rewards

Training-data presence

For older or general queries, ChatGPT often answers without retrieval at all, drawing on training data. This rewards brands that have been heavily covered in trade press, Wikipedia and large public corpora over years. New entrants, even with strong current visibility, take longer to register here.

Cross-source corroboration

ChatGPT’s citation behaviour suggests it weights claims that appear across multiple training sources. A fleet size mentioned only on your own website is a weak signal; the same number repeated across three trade press articles, your Wikipedia entry and an industry directory becomes confident enough to quote without hedging.

Brand consistency across modalities

ChatGPT’s training corpus included social, blog, press and structured data. Brands whose presentation is consistent across all of these (same legal name, same fleet count, same office locations) score better than brands whose digital presence has drifted across channels.

OAI-SearchBot and ChatGPT-User retrieval

For browsing-enabled sessions, ChatGPT does its own retrieval through OpenAI’s crawlers. The retrieval is generally less aggressive than Perplexity’s: fewer sources cited per answer, more conservative phrasing.

The practical implication

If you are starting from a weak position in both, the work order is roughly:

1. Fix structural quality and schema first. This helps everywhere.
2. Concentrate on factual consistency across sources. This helps ChatGPT more, but hurts no one.
3. Publish regularly and maintain page freshness. This helps Perplexity more.
4. Earn tier-one trade press features. This helps both, with a longer lead time on ChatGPT.
5. Track the engines separately in your quarterly audit. The trends will diverge and the divergence is informative.

Where the engines agree

Both reward authority, both reward specificity, both reward schema-clean structure. The brands that are invisible in both are doing all three things badly. The brands that are visible in one but not the other usually have a specific gap: heavy backlog but no recent publishing (visible in ChatGPT, invisible in Perplexity), or strong current activity but no historical trade press depth (visible in Perplexity, invisible in ChatGPT).

The fix is rarely to overoptimise for one engine. It is to fix the gap that explains the divergence. Once the gap closes, performance improves in both.

A note on Gemini and Copilot

Gemini behaves closer to classical Google search in citation logic and rewards what Google rewards: depth, internal linking, E-E-A-T. Copilot behaves closer to Bing-style retrieval and overlaps heavily with Perplexity’s preferences. For most maritime brands, optimising for ChatGPT and Perplexity also covers Gemini and Copilot adequately. Engine-specific tuning matters more in highly competitive categories where small differences swing the citation set.