Key takeaways:
- Google’s AI has a fixed grounding budget per query. More content does not increase what AI can read.
- Rank higher, get seen more by AI. Top positions receive a larger share of the grounding budget.
- AI reads in snippets, not pages. Only select, answer-dense sections are extracted.
- Long content dilutes AI visibility. Beyond a point, added length reduces usable coverage.
- Density beats length. Clear, focused sections outperform sprawling pages.
- Structure enables selection. Well-defined sections are easier for AI to extract.
- SEO still gates AI exposure. Rankings now control AI visibility, not just clicks.
Introduction
For years, SEO has rewarded one dominant strategy: write more comprehensive content than everyone else. Long-form guides, pillar pages, and mega articles became the default playbook for ranking and visibility.
AI-powered search changes that equation entirely.
Google’s AI does not evaluate your content the way humans do, and it does not read your pages end-to-end.
Instead, it operates within a grounding budget: a fixed amount of content it can reference when generating AI answers. Once that budget is exhausted, nothing else on the page matters.
Understanding this constraint is now essential for anyone optimising content for AI-driven search experiences.
What Is a Grounding Budget?
A grounding budget is the maximum amount of content Google’s AI is allowed to reference when generating an answer to a single search query.
Instead of reading entire webpages, Google’s AI works within a fixed window of text. From that window, it selects short, relevant passages from a small number of sources and uses them as the factual backbone for AI-generated responses in experiences like AI Overviews and AI Mode.
A simple way to think about it is this:
Google’s AI has a limited “reading allowance” per question. Once that allowance is used up, everything else on the page, no matter how well written, is effectively invisible to the AI.
This is why adding more words to a page doesn’t help beyond a point. A 4,000-word article doesn’t give the AI more room to read than a 1,200-word one. It simply means your content is competing with itself for inclusion inside the same fixed window.
Now zoom out one level further.
That grounding budget isn’t given to just one page. It’s shared across multiple results.
Higher-ranking pages receive a larger portion of that window, while lower-ranking pages receive less. So you’re not only competing against other sites; you’re competing for how much of your content is allowed into the AI’s context at all.
In practical terms, this means you’re not trying to make Google read more.
You’re trying to make sure the right parts of your page are chosen within a limited, highly competitive space.
What the Data Reveals About Google’s Grounding Budget
To move beyond theory, it’s important to anchor the idea of a grounding budget in real data.
In “How big are Google’s grounding chunks?”, Dan Petrovic analysed 7,060 real search queries and compared the exact text snippets supplied to Google’s Gemini systems against 2,275 tokenized webpages. This allowed him to observe how much content Google’s AI actually consumes, and how that content is allocated.
The most important findings are:
- Google’s AI operates with a fixed grounding budget of roughly 2,000 words per query. Across thousands of searches, the median total amount of text supplied to the model was around 2,000 words, regardless of how long the source pages were or how many sources were involved. Adding more content or more pages did not increase this total, as it was simply redistributed.
- Ranking position determines how much of that budget a page receives. Pages ranking at the top consistently received the largest share of the grounding budget. The #1 result alone accounted for roughly 28% of the total grounded text, while the #5 result received about 13%. In practical terms, a top-ranked page received about twice as much AI-visible content as a lower-ranked competitor.
- Only a few hundred words are typically taken from any single page. At the page level, the typical source contributed around 377 words to the AI’s context. Nearly 77% of pages had between 200 and 600 words selected, even when the original page was several thousand words long.
- There is a clear ceiling on per-page extraction. The amount of content selected from a single page plateaued at around 540 words. Beyond this point, adding more text to a page did not meaningfully increase how much the AI used.
- Coverage drops sharply as pages get longer. Shorter pages often had a majority of their content represented in AI grounding, while very long pages saw only a small fraction used. Pages under 1,000 words could see well over half their content selected, whereas pages exceeding 3,000 words often had less than 15% coverage.
Taken together, these figures reveal a hard constraint: AI visibility is capped. Once the grounding budget is exhausted, additional content no longer contributes to AI answers. It simply dilutes the proportion of your page that can be surfaced.
As this data shows, success in AI search is no longer about how much you write, but how much of what you write earns a place inside a very limited window.
Why Rankings Still Control AI Visibility
One of the most important consequences of a grounding budget is that traditional rankings remain a gatekeeper for AI visibility.
When multiple pages are eligible to be referenced, Google’s AI does not divide its attention equally. Pages that rank higher organically receive priority access to the grounding budget, while lower-ranking pages receive progressively less representation.
This means:
- Ranking higher doesn’t just increase clicks. It increases how much of your content the AI is even allowed to see.
- Poor rankings severely limit your presence in AI-generated answers.
AI search has not replaced SEO fundamentals. It has amplified their impact.
How Google Actually Uses Your Page
Another critical shift is how pages are evaluated.
Google’s AI does not treat your article as a single unit. Instead, it scans for answer-dense segments that can be extracted and reused. These segments are often small, self-contained blocks of information that directly satisfy a specific intent.
As a result:
- Large portions of long pages are never considered.
- Only select sections influence AI responses.
- Content buried deep in long narratives often goes unused.
This explains why adding more sections, examples, or commentary beyond a certain point yields diminishing returns for AI visibility.
Why Longer Content Loses Its Advantage
In traditional SEO, long content worked because it covered more subtopics, attracted more links, and matched broader intent.
In AI search, length becomes a liability.
Because the grounding budget is capped:
- Additional words do not increase AI usage.
- Coverage of your page decreases as the length increases.
- Only a shrinking fraction of your content is eligible to be surfaced.
At scale, this creates a paradox: The more content you add, the less of it the AI can practically use.
This does not mean long content is “bad”. But it does mean length alone no longer creates an advantage in AI-mediated visibility.
Information Density Is the New Ranking Multiplier
The real winner in AI search is information density.
Pages that perform best tend to:
- Focus on a single, well-defined intent.
- Deliver clear answers quickly.
- Use structured sections with obvious topical boundaries.
- Avoid unnecessary padding or narrative filler.
A tightly written page where every section earns its place is far more likely to be extracted, referenced, and reused by AI systems than a sprawling article where only fragments are relevant.
In AI search, clarity beats comprehensiveness.
What This Changes for Content Strategy
The grounding budget introduces a fundamental shift in how content should be planned and written:
The goal is no longer to be the longest page on the topic. The goal is to be the most extractable.
Each section should stand on its own as a potential AI-ready answer. Rankings determine how much visibility your content can even compete for
This pushes content strategy toward:
- Intent-focused pages instead of bloated hubs.
- Strong internal segmentation rather than endless scrolling.
- Precision over verbosity.
AI search, therefore, rewards pages that respect its constraints.
Conclusion: The New Mental Model for AI-First Content
Instead of asking: “How do we make this article more comprehensive?”
The better question is: “Which parts of this page deserve to be selected?”
Because in AI search, visibility is not about how much you publish, but about how much of what you publish survives the grounding budget.
As AI search continues to evolve, content strategy must shift from publishing more to engineering relevance. The future belongs to pages built deliberately for extraction, where every section earns its place inside the grounding budget.