Generative engine optimisation for WordPress

What GEO means

Generative engine optimisation is the work of making an AI assistant describe you accurately and cite you when it answers a question in your field. The unit of success is not a position in a list. It is whether your name appears in a paragraph somebody reads instead of visiting a search page at all.

It splits into two jobs that get conflated constantly. Being retrievable means the lookup step can reach your page. Being quotable means that once it has your page, your sentence is the one it uses. Most advice covers the first and skips the second, which is why sites get crawled regularly and still never appear in an answer.

How an assistant actually answers

When an assistant answers something current, it usually does three things in sequence. It runs a search to find candidate pages. It fetches some of them and reads a portion of each — a chunk, not the whole document. Then it writes an answer from what it read, citing some of the sources.

Every one of those steps is a place you can be eliminated. A robots.txt rule removes you at step one. A page that needs JavaScript to render removes you at step two. Vague, hedging prose removes you at step three, because there is no sentence worth quoting.

This is why GEO work feels scattered: it is three different problems wearing one name.

The crawlers, and what blocking each one costs

The mistake almost everyone makes is treating these as one switch. A single company runs several agents doing different jobs, and the consequences of blocking them are not remotely the same.

Vendors add, rename and retire agents, so treat this as a starting point and check the current documentation before committing a policy. Two further points: directives are matched per user agent, so a group you do not name falls through to your User-agent: * rules; and robots.txt is a request rather than a control. Named crawlers from large vendors honour it, anonymous scrapers do not, and nothing in a text file protects content that should not be public.

A common position is to allow the search and user-fetch agents while declining the training ones. That keeps you citable today and out of the next training run, which is what most people actually want when they say they want to "block AI".

Tell machines what you are, rather than making them infer it

Two pieces of structured work do most of the heavy lifting, and neither is a ranking trick. Both exist to stop a model guessing.

• Organization schema on your home page: name, URL, logo, a one-line description, and sameAs pointing at profiles that describe the same entity elsewhere. This is the anchor everything else hangs from.
• Article schema on anything written: headline, author, published and modified dates. The dates matter more than people expect.
• FAQPage schema where you genuinely have recurring questions, with every question and answer visible on the page. Marking up a FAQ the reader cannot see is against the guidelines.
• One canonical description of yourself, word for word, everywhere. Most sites describe themselves four different ways — homepage, About, footer, meta description — and a model assembling a description has to pick one, possibly your weakest.
• One consistent name. If a product name, an abbreviation and a domain appear interchangeably without ever being connected, they can be read as several different things, each holding a fragment of the evidence.

The rule that governs all of it: structured data must describe what is actually on the page. A model that finds markup contradicting your visible text has learned your markup is unreliable and cannot tell which half to trust. You are better off with no schema than with schema that disagrees with your own page.

How to measure it

There is no Search Console for this. You have three imperfect signals, and it is worth being clear about what each one can and cannot tell you.

• Ask, on a schedule. Put your customers' five real questions to the assistants monthly and write down the answers and the citations. Crude, manual, and the only direct measure that exists.
• Referral traffic. Visits with an assistant domain as referrer are real but undercount badly, because most people read the answer without clicking.
• Crawler hits in your access log. These prove you are reachable. They prove nothing whatsoever about being cited.

Start with the first one before changing anything, so you have a baseline. Fifteen minutes now saves you guessing later about whether a change did anything.

Where the plugin fits, and where it does not

We will not tell you a plugin makes you appear in AI answers, because no one controls that and the honest version is more useful to you.

What WP Content Writer Pro contributes is the content half: drafts with heading structure that works as chunk boundaries, sections written to stand on their own, and a fact-checking pass that flags claims it could not verify instead of smoothing them into confident prose. That last one matters here more than anywhere else — a confident wrong sentence is exactly the thing that gets quoted.

Crawler policy, schema and entity consistency are site-level jobs. They are not hard, they are mostly done once, and no plugin should claim to have done them for you.

Further reading: what GEO is, the crawlers in detail, and schema markup for AI search.

Questions