Is GEO Worth Doing on a Small Site?

On a small site, three free changes carry most of the benefit available, and almost everything sold under the GEO label carries the rest — badly, per dollar. Publish pages that answer one question each in crawlable HTML, decide deliberately whether AI crawlers may read them, and hand-check a short list of prompts once a month. That is the whole small-site program, and it costs editing time.

One disclosure first, because it changes how you should read the rest: our own evidence comes from three large production builds. We have no small-site case study. The reasoning below is drawn from platform documentation and published click studies, and where our fleet has nothing to say, this page says so.

What should a small site actually do first?

Three things, in this order, none of which require buying anything.

Make every page answer one question, early. Put a direct answer of roughly 40–75 words at the top, phrase your headings as the questions people type, and keep paragraphs short enough to be quoted whole. This is the part you fully control, it helps ordinary search at the same time, and it is free.

Ship that answer in HTML that arrives without JavaScript. A crawler gets whatever your server returns to its request. If your answer only exists after a client-side render, it is not in that response. Run curl against your own URL and read what comes back — if the answer text is not there, fix that before anything else on this page.

Decide about AI crawlers on purpose. Allow or disallow GPTBot, ClaudeBot, PerplexityBot and Google-Extended because you chose to, not because a default did. For most small sites that want to be findable, the answer is allow — but it is a decision with tradeoffs, not a checkbox.

Then, monthly: run the same 10–20 questions through two or three assistants by hand and write down what they say and who they cite. The method is measuring AI share of voice without paid tools.

What is the realistic upside?

Small, and the published numbers agree on the direction if not the magnitude. Stamp the periods, because these are moving figures:

Read together: the surface takes clicks away from everyone, and being the cited source claws part of it back. That is worth having and it is not a growth channel by itself. Why the 1% and the +35% are both true — different denominators, different questions — is worked through in are AI citations worth anything.

For a site with a few hundred monthly sessions, a percentage of a percentage is a handful of visits. The honest reason to do the free three anyway is that they are the same changes that make a page better in ordinary search, and their cost is an afternoon.

Should a small site pay for AI visibility tracking?

Almost never, and this is the recommendation most against our own interest, because tooling and retainers are what this niche sells. A subscription measures something you cannot yet act on: at small scale you will have too few citations for a trend line to mean anything, and the tool's sampling is not deterministic, so month-to-month movement can be noise.

The substitute is a spreadsheet. Ten to twenty questions, run by hand, once a month, logged with dates and screenshots. It takes about an hour, it produces the same directional signal, and it forces you to read the answers — which is where the useful information actually is. What the paid tools cost, and what they add when you outgrow the spreadsheet, is in AI visibility tool pricing.

The same logic applies one level up: a small site that cannot fund maintenance should not fund an agency either. The three-way comparison, including when hiring anyone is the wrong call, is in-house vs agency vs tools.

Which small sites should skip this entirely?

Three kinds, plainly.

Sites whose demand is branded or transactional. If people search your name or search to buy, an AI summary sits further from the money than your product pages do. Fix the product pages.

Local walk-in businesses. Demand that depends on physical proximity is mostly served by map and listing surfaces rather than by a library of answer pages, so the effort usually belongs there instead. We run no local build, so treat that as a judgment call rather than something we have measured.

Sites with no capacity to maintain what they publish. A page with a stale figure is a liability on every surface at once — an assistant quoting your outdated number attributes it to you. If maintaining twenty pages is already hard, publishing forty is not the answer.

What does "small" actually change about the work?

The economics, not the technique. Everything a large library does — one question per URL, answers at the top, sourced figures, crawlable HTML — is identical on a ten-page site and free to apply. What changes is that the machinery around it stops paying: automated gates, tranche briefs, parallel production and freshness enforcement are answers to a coordination problem a ten-page site does not have. One careful person and a checklist cover it.

Google's own guidance for generative features points the same way: useful, crawlable, people-first content, with no separate technique reserved for AI surfaces. That is the whole small-site advantage — nothing on the free list is gated behind budget.

And the limit is worth restating, since nobody selling this says it: none of this promises citations, rankings or traffic. It makes your pages eligible to be retrieved and quotable when they are. If you want the same reasoning at full depth, with the evidence tiers attached, it is our generative engine optimization guide.