What AI changed: content, search, research · Content and search

Is GEO worth a quarter of the content budget

Decide whether a $10,000-a-quarter GEO retainer gets a quarter of the content budget, from its ceiling, and set the measures that would justify it.

casemediumExercise · 3 min · 41 of 60

On this page

An agency wants a quarter of the content budget to make your pages more citable by answer engines. The case turns on a ceiling you can compute today and an effect nobody has measured yet.

The situation

A software company gets 20,000 organic sessions a month, and analytics attributes 120 of them, 0.6%, to answer-engine referrers. Twenty of its pages are already cited in answers. The content budget is $40,000 a quarter, about 25 written posts at $400 plus distribution. An agency proposes taking $10,000 of it, a quarter, for a GEO retainer: rewrite the 20 cited pages with added quotations, statistics and cited sources, which it says "raises visibility by up to 40%", quoting the KDD 2024 paper. Its forecast is 30 more sessions a month per page. Assume an organic session is worth a third of a $2 paid search click, about $0.67. Branded search runs at 3,000 queries a month, flat for a year. Nobody has measured citation share. An analyst's hour costs $100.

The question

Decide whether GEO gets a quarter of the content budget this quarter, and say what you measure before next quarter's decision.

A strong answer · Rubric · Where answers go wrong

Write your answer first, then open

A strong answer

Not this quarter. The ceiling first: 20 pages × 30 sessions is 600 a month, 3% of organic. At $0.67 that is $400 a month, $1,200 a quarter, against $10,000. The retainer pays only if an answer-engine session is worth $10,000 ÷ 1,800 ≈ $5.56, more than eight times an organic one.

The 40% does not close the gap. In the paper, visibility is the share of a generated answer attributed to a page on a benchmark, not sessions. And citations are rarely clicked: Pew's panel clicked a link inside an AI summary on 1% of visits. If GEO pays, it pays as presence without a click, which only citation share and branded search can show, and neither has a baseline.

So spend about $2,500, 25 analyst hours, on measurement and keep the rest on posts:

citation share: 50 buyer questions, each run 5 times a month, 250 answers, read with its standard error of about 2 points;
AI-surface referrals monthly, as a floor, since some arrive without a referrer;
branded search, which cited-but-unclicked answers would move.

Then offer the agency a paid test instead of a retainer: rewrite 10 of the 20 cited pages chosen at random, leave 10 untouched, and compare the halves after a quarter.

What would change the decision: rewritten pages beating the untouched ones on citation share by more than about 4 points, with branded search rising off its flat 3,000; or AI-surface referrals growing past about 5% of organic, which would make the ceiling large enough to fund.

Rubric

computes the ceiling (600 sessions, 3% of organic) and values it against the $10,000, with the break-even value per session
separates the paper's visibility metric from sessions, and uses the evidence on how rarely citations are clicked
names citation share, AI-surface referrals and branded search, with how each is counted and how noisy it is
proposes a controlled test, rewritten pages against untouched ones, with a duration
states the result that would reverse the decision
neither prescribes GEO tactics nor rejects them before measuring

Where answers go wrong

"Yes: answer engines are where search is going, and 40% more visibility is worth a quarter of the budget." It treats a benchmark's visibility metric as traffic and never computes the $1,200 ceiling, so it spends $10,000 on a claim it cannot check.

"No. Google says AI features need no special optimisation, so GEO is a scam." The platform owner's guidance is about its own engine and is not neutral, and the answer refuses to measure, which leaves the company blind if referrals grow. The lesson asks for a measurement, not a verdict.

"Run the retainer for a quarter and see whether sessions rise." With no untouched pages to compare and no baseline for citation share or branded search, any seasonal or engine change in that quarter will be read as the agency's effect.