How to Optimize for Perplexity and ChatGPT Search

AI search engines cite pages Google doesn't rank. Here's how to be the source they pick. This piece walks through the exact approach we use with clients, and the mistakes we've learnt to avoid along the way.

Why this matters in 2026

Search has moved on from the tactics that worked five years ago. How to Optimize for Perplexity and ChatGPT Search is one of the areas where the old playbooks either quietly fail or actively backfire. The upside is real, but only if the work is sequenced correctly.

The core principle

Every strong system here comes back to one idea: make it obvious to Google (and to a human skim-reader) what the page is for, who it's for, and why it deserves to rank. Anything that doesn't serve that goal is decoration.

The step-by-step process

Start by auditing what you already have. Map current coverage against the opportunity, then decide whether the right move is to expand, merge, prune, or rebuild. Doing this before you write a single new page saves weeks of wasted effort.

Next, build the smallest useful version. Ship it, measure it against a real baseline, and only then invest in scale. Most teams over-invest in production and under-invest in feedback loops.

What we track

Impressions, clicks, non-brand traffic, and assisted conversions — in that order. Rankings are a leading indicator, but revenue is the only metric a CFO will renew a budget for. Wire both into the same dashboard from day one.

Common mistakes to avoid

Rushing to publish before the topic map is stable. Treating one big update as "done" rather than a rolling programme. Optimising for a metric no one downstream cares about. Copying a competitor's structure without understanding why it works for them.

How to get started this week

Pick one page, one query, and one measurable outcome. Ship a small change, wait two weeks, and read the result honestly. That single loop, repeated, is worth more than any 90-day plan you'll write on a whiteboard.

If you want tooling that plugs into this workflow, take a look at the Zeshly platform — it's built for exactly this kind of iteration.