Short answer. For commercial and evaluation-stage questions, 83% of AI citations came from pages updated within the previous twelve months and over 60% from pages refreshed within six. For most organisations that makes updating the cheapest available intervention — and the one most likely to be done badly, because re-dating a page is easier than changing it. A refresh programme reporting "60 pages updated this quarter" with no record of what changed in them has reported a date field.
Refreshing beats publishing on the evidence, on cost, and on how quickly it can be verified. But the same volatility that makes AI answers hard to influence also makes a refresh hard to prove, and most refresh programmes lose credibility at exactly that point.
This piece sets out what counts as a refresh, how to run one that holds up, what cadence the evidence supports, and how to demonstrate it worked against a noisy baseline.
Why does refreshing beat publishing?
Three findings, taken together, point the same way.
| Finding | Figure | What it implies |
|---|---|---|
| Freshness on commercial queries | 83% of citations from pages updated within 12 months; 60%+ within six | Recency is heavily rewarded |
| Position within the page | 55% of sampled AI Overview citations came from the first 30% of the cited page | Front-loading the answer is the highest-leverage structural edit |
| What kind of edit works | Authority-style edits — adding citations, statistics and quotations — raised visibility by up to 40%, outperforming rewriting and simplification; keyword stuffing performed worse than no change at all | The winning edits are additive and factual, not stylistic |
The third row comes from Aggarwal et al., "GEO: Generative Engine Optimization" (ACM SIGKDD 2024), benchmarked across roughly 10,000 queries and nine datasets.
Every one of those edits is cheaper to make on a page that exists than on a page that does not. A thorough page from 2023 that nobody has touched competes badly against a thinner page revised last month.
What actually counts as a refresh?
The distinction that matters is between changing the date and changing the page.
| Level | What happens | Effect | Honest? |
|---|---|---|---|
| Re-dating | The published date changes; nothing else does | None, and it misleads readers | No |
| Cosmetic rewrite | Prose is smoothed, headings reworded | Minimal — this is the category the GEO benchmark found underperforms | Yes, but low value |
| Fact refresh | Every figure re-verified, stale ones replaced or retired, new sources added | The intervention with evidence behind it | Yes |
| Structural refresh | Answers moved to the front, headings rephrased as questions, tables added | Addresses the position-in-page and extractability findings | Yes |
| Scope refresh | New sub-questions added to cover the fan-out neighbourhood | Enters more retrieval pools | Yes |
The last three are the work. The first two are what most quarterly refresh reports are actually counting.
What does a refresh operation that holds up look like?
- Build a claims inventory before touching anything. Every page, every figure on it, the source behind each figure, and the date of that source. Without this, "re-verify" has no object, and this is the step everyone skips.
- Sort by decay risk, not by traffic. A page carrying a 2024 statistic in a fast-moving category is higher risk than a stable explainer with more sessions.
- Re-verify against the original source, not a secondary. Aggregator posts recycle figures long after the underlying study is superseded, and per-tactic percentages circulate that do not appear in the papers they are attributed to.
- Retire what you cannot stand behind. Removal is a legitimate outcome and the rarest one in practice. A wrong claim that keeps resurfacing is worse than a missing one.
- Fix the top 30% of the page first. More than half of sampled citations come from there.
- Add the missing sub-questions. Surfer SEO's December 2025 analysis found pages ranking for a main query plus at least one fan-out query were 161% more likely to be cited than pages ranking for the main query alone.
- Update the machine-readable dates honestly, and only when something changed.
- Record what changed, by whom, on what date. This is the audit trail, and it is also the only way to attribute a later movement to anything.
Step one is what makes the rest possible. Holding figures in a single registry keyed by source — so a page references a key rather than restating a number — means a statistic cannot appear without a source attached, and updating a source updates every page that used it.
How often should pages be refreshed?
There is no universal interval, but the evidence bounds it.
- Twice a year, with real changes, is a defensible baseline for pages that answer buyer questions. It sits comfortably inside the six-month window that carried over 60% of commercial-query citations.
- Quarterly for anything carrying a fast-moving figure. Market shares, pricing, platform behaviour and tool capabilities in this category all moved materially within single quarters during 2025–2026.
- On event, always. A product change, a pricing change, a leadership change or a superseded study is a trigger regardless of the calendar.
- Annually is the floor. Past twelve months a page falls outside the window that carried 83% of commercial-query citations.
How do you prove a refresh worked?
This is where refresh programmes lose credibility, because the measurement environment is hostile. Parse, analysing 693,509 answers between March and April 2026, found that asking the same question twice returned only 21.2% of the same cited domains on ChatGPT and 31.5% on Google AI Overviews. Within a one-week window, overlap rose only to 26.7% and 36.8%.
Against that noise floor, a before-and-after screenshot proves nothing. What does hold up:
- Measure a rate across repeated runs, before and after, on a frozen question set — not a position, not a single check.
- Keep an unrefreshed control group. Comparable pages you deliberately do not touch in the same period. Without one, you cannot separate your refresh from the models changing.
- Allow weeks, not days. Retrieval surfaces respond in days to weeks; proving it against the noise takes longer than the effect does.
- Expect nothing on memory mode. Answers with search off change when a model is retrained, whatever you republish.
- Report the control alongside the treatment. A refresh programme that reports only treated pages is reporting the market, not its own work.
The unrefreshed control group is the single cheapest credibility upgrade available to a content team, and almost nobody keeps one, because it feels like deliberately neglecting pages. It is the difference between "citations rose four points" and "citations rose four points against a control that moved one".
Limits of the evidence
- The freshness figures are compilations, directionally consistent but not a single controlled study.
- Recency is a signal, not a mechanism. Updating a page that answers nothing does not make it citable.
- Refresh cannot reach the 85%. Roughly 85% of AI references point at third-party sources, and none of them are on your publishing schedule.
- Some decay is not fixable by editing. A superseded study should be removed rather than updated, and the claim it supported may simply no longer be available to you.
How Lifewood approaches this
Lifewood holds figures in a single registry keyed by source, so a page references a key rather than restating a number. That is a forcing function rather than a convenience: a statistic cannot appear on a page without a source attached, and updating a source updates every page that used it. It is also why the same figures recur across articles with identical wording rather than drifting.
Refresh work is sorted by decay risk rather than by traffic, re-verified against the original source rather than a secondary, and logged as claim, source, publisher, date and reviewer at the point of editing rather than reconstructed afterwards. Removal is treated as a legitimate outcome and recorded as one.
Every refresh cycle holds an untouched control group of comparable pages, reported alongside the treated set including when the two moved together. The measurement runs on a frozen question set with retrieval and memory surfaces kept apart, because a refresh cannot move memory mode and reporting them blended makes correct work look like failure. See what gets you cited by AI answer engines and how Google AI Overviews picks sources.
Sources and further reading
- Omnibound, Answer Engine Optimization statistics 2026 — the freshness, position-in-page and third-party share figures.
- Aggarwal, Murahari, Rajpurohit, Kalyan, Narasimhan & Deshpande, GEO: Generative Engine Optimization, ACM SIGKDD 2024.
- Surfer SEO, AI Overview fan-out rankings boost citation odds by 161%, December 2025, via Search Engine Land.
- Parse, AI citation volatility by industry, 693,509 answers, March–April 2026.