Why not just optimize for all of them?
You largely do. The fundamentals — crawlable HTML, answer-first structure, sourced claims, clean structured data — help with every engine, so most of your effort isn't engine-specific. Prioritization mostly affects where you measure and which citation behaviors you study, not whether you have to write different content for each.
How do they differ?
They differ in reach and in how visibly they cite. Some surface sources prominently and lean on live retrieval; others answer more from the model with lighter attribution. Some are tied to a search index, others to their own crawl. These differences shape how you track citations on each, but they rarely change the underlying content work.
Together they account for the majority of AI-cited traffic today. Build your baseline citation tracking there before spreading to other engines.
How should I prioritize?
By where your buyers are. A developer tool and a local clinic have different audiences using different assistants, so ask (and check analytics and referral data for) which assistants actually send you people. Weight your measurement toward the engines that matter to your market, and don't burn effort chasing one that your customers never open.
What about new or changing engines?
The lineup shifts, and engines change how they retrieve and cite on their own schedule. Because your core work is engine-agnostic, you adapt by adjusting what you monitor rather than rebuilding your content. Build on fundamentals and you're resilient to the churn.
The short version
Follow your audience: usually ChatGPT and Google AI Overviews for reach, Perplexity and Claude alongside. Do the engine-agnostic work once, then measure per engine where your buyers actually are.