Four factors drive that decision:
The page that creates the least uncertainty usually wins.
What We Call Relevance Arbitration
Two pages can cover the same topic, both well-written and thorough, and still land far apart in the results. One ranks at the top consistently. The other barely shows up. This is not a quality problem. It is an uncertainty problem.
Google cannot simply declare one page "better" than a similar competitor. It has to determine which page it can most confidently recommend, and it has to do that reliably, across billions of searches, every day.
We call this process relevance arbitration. This is not an official Google term. No public Google documentation names it this way. It is our name for a real, observable pattern in how Google's ranking system behaves under uncertainty, based on Google's own public statements about ranking, sworn testimony from Google's engineers, and the leaked internal documentation covered in Stage 6 below.
How it works, at a high level:
multiple pages could all answer the same search. Google evaluates a set of confidence signals for each one. Uncertainty decreases as those signals accumulate. One page becomes the safest recommendation.
Google is not optimizing for which page is best. It is optimizing for which page it can select with the most confidence. Those two things are related, but they are not the same thing. The winner is not the most impressive page. It is the page that gives Google the least reason to doubt it.
Relevance arbitration has six stages:
"Most SEO advice treats ranking like a single scoring system: write the best page and the algorithm hands you the top spot. That's not what we see when we audit client sites. Two genuinely strong pages compete for the same position constantly, and the one that wins usually isn't the more impressive one. It's the one Google has the least reason to doubt."
Davit Nazaretyan, Founder of LinkyJuice
Stage 1: Eligibility (Getting Into the Running)
Before any comparison happens, Google determines who is even in the race.
Google reads pages, groups them by meaning, and builds a pool of candidates for each search. Only pages close enough to the topic enter that pool. Only pages inside the pool get compared to each other.
Two pages on the same topic are not flagged as duplicates just because they cover similar ground. They compete for the same position because Google treats them as close-enough alternatives to each other.
This stage evaluates topic proximity, not quality. Two pages can differ wildly in depth and still compete head-to-head, simply because they are close enough in meaning. Once two pages share a pool, arbitration begins.
Stage 2: Relevance Filtering (Does This Page Actually Fit?)
Relevance filtering asks whether a page fits the search, not just the keyword.
Google looks past the words in a query to identify what the searcher actually needs. A search for "how to do something" requires different content than a search for "why something happens," even when the topic is identical. Same keyword, different intent, different answer required.
This is what relevance actually means in SEO: intent alignment, not keyword matching.
Google asks whether a page was built to answer the kind of question being asked. A page that fits the intent precisely enters the next stage carrying less uncertainty. A page that only partially fits carries more. This stage does not crown a winner. It sets the starting conditions for every stage after it.
Stage 3: Quality Threshold (Is This Actually Useful?)
The quality threshold checks whether content is genuinely useful or only surface-level.
This stage functions as a gate, not a ranking signal by itself. Pages that clear it stay in contention. Pages that do not fall out of contention entirely.
Real substance clears the gate: original information, genuine depth, evidence that someone who understands the topic wrote the page. A page that only restates what every other page already says gives Google no reason to prefer it, which is its own form of uncertainty.
This stage is also where experience and authoritativeness matter. Google assesses whether content reflects real knowledge or a competent reshuffling of existing content. A page that adds something new, takes a clear position, or shows firsthand understanding reduces uncertainty. A page that plays it safe and generic increases it.
In close competitions, both pages often clear this threshold. When they do, arbitration moves on to trust.
Stage 4: Trust Calibration (Who Vouches for This Page?)
Trust calibration is where close competitions actually start to separate.
Google ranking depends on both backlinks and relevance, applied at different stages. Relevance gets a page into contention. Trust is often what tips the final decision.
Google looks at which page has more credible external endorsement, mainly through links from trusted sites. This is not a link-counting exercise. It measures trust signals. When credible sources link to a page, they give Google more confidence in recommending it.
Even a small trust difference can shift the arbitration. The page with slightly stronger trust signals becomes the safer choice. The other page stays valid, just less certain. For more on evaluating which backlinks actually carry this kind of weight, see how to know if a link is actually worth getting.
This is the stage where arbitration stops filtering and starts producing real divergence between two pages.
Stage 5: Context Validation (Does This Site Make Sense Here?)
Context validation asks whether a page belongs to a site consistently associated with its topic.
A strong page on an unfamiliar site carries more uncertainty than the same strong page on a site Google already connects with the subject. Site-level context acts as a confidence multiplier. It either backs up a page's position in the arbitration, or quietly introduces doubt about whether the page is a reliable source.
This is not about domain authority as a single metric. It is about whether the site surrounding the page makes it easier or harder for Google to commit to recommending it.
A single strong page surrounded by unrelated content is harder for Google to trust than the same strong page inside a site clearly built around that topic. For more on building that kind of site-level consistency, see how to build a strong backlink profile.
Stage 6: Behavioral Reinforcement (What Do Users Actually Do?)
After Google makes an initial selection, it watches what users do next.
This stage used to be one of the more disputed claims in SEO. Google representatives publicly denied for years that click and engagement data directly influenced rankings. Two independent, verifiable sources changed that. Pandu Nayak, Google's VP of Search, testified under oath during the 2023 DOJ antitrust trial that a click-based re-ranking system called NavBoost is "one of Google's strongest ranking signals." The May 2024 Content API Warehouse leak subsequently named the specific fields NavBoost tracks, including goodClicks, badClicks, and lastLongestClicks, inside Google's internal production code.
When people click a result and stay, that confirms the selection was right. When they bounce straight back to the results page, that signals it was wrong. This data feeds back into how confidently Google holds a page's position, aggregated over a rolling window of roughly 13 months.
This explains why a page's ranking can drift even when nothing about the page changed. User behavior continuously reinforces or erodes Google's confidence in its own decision.
A page that is technically strong but consistently disappoints users loses ground over time. A page that does not look exceptional on paper but keeps people satisfied holds its position. This is a feedback loop: positive behavior locks in the selection, negative behavior reintroduces the uncertainty Google was trying to eliminate.
Why Small Differences Turn Into Big Gaps
Two similar pages can trade positions early on, while arbitration has not yet settled between them.
The process compounds after that. Once one page accumulates a small advantage at one stage, that advantage feeds into the next stage. A slight trust edge leads to more consistent selection. More consistent selection produces better behavioral signals. Better behavioral signals reinforce the trust. The gap widens.
This is why a page that started slightly ahead can end up miles ahead, and why a page that started slightly behind can struggle to catch up even after real improvements. Google did not make a mistake and forget to correct it. The page with an early lead built up enough compounding confidence that the gap became structural. Closing a structural gap means overcoming a reinforcement pattern, not just a quality difference.
The Mental Model That Changes How You Think About SEO
Most SEO advice treats ranking like a quality competition: write the best page, optimize it well, and it should win.
That model is incomplete. It is the reason good pages sometimes lose to pages that seem objectively worse.
Google is not converging on the best page. It is converging on the most confidently selectable page under uncertainty. Those are not the same thing.
The mental model shift:
Old SEO ThinkingBetter ModelGoogle finds the best pageGoogle selects the safest answerRankings are a simple scoreRankings reflect confidenceQuality automatically winsSmall advantages compoundOptimization means adding signalsOptimization means reducing uncertainty
A page can be excellent and still carry unresolved uncertainty at the trust stage, the context stage, or the behavioral stage. Any one of those is enough to tip arbitration toward a competitor.
The real question is never "is this page good enough." It is "does this page give Google anything to doubt."
Remove the doubt at every stage, and the ranking follows. If you are working through this on your own site, our SaaS link building services can help you see how your pages are actually competing inside Google's system. Book a call with us.
About the Author
Davit Nazaretyan writes about link building, backlink audits, and SEO automation at LinkyJuice. His work focuses on separating what actually moves rankings from what just looks like it does: from cleaning up toxic backlink profiles to building outreach systems that hold up against algorithm updates. Connect on LinkedIn.
