Two AI-search tips can still be one piece of evidence

What we actually reviewed

In a review of 15 saved Instagram reels, two SEO videos described the same story about a LinkedIn post appearing in an AI answer. We grouped them as one underlying anecdote. The posts were separate; the reported example was shared.

The review used timestamped machine transcripts and sampled video frames. It was not a controlled search experiment, and we did not independently reproduce the claimed ranking behavior. The observation is narrow: those two videos do not provide two independent demonstrations.

Separate observation from explanation

A screenshot can show an answer at one moment. It cannot by itself establish how often that answer appeared, whether other users saw it, or which change caused it. A useful review preserves the screenshot as an observation while keeping the causal explanation open.

The same distinction applies to advice about putting “Reddit” into a title or URL. A third reel in the review proposed this tactic but supplied no reproducible test. That is a hypothesis to investigate, not a reason to add irrelevant words to pages.

Use an evidence ledger

Keep the smallest useful record for each claim. When another post repeats the same source, attach it to the existing record instead of increasing the evidence count.

A fair evaluation should also preserve cases where the tactic fails. Otherwise a collection of memorable examples can become a misleading picture of reliability.

• Claim: write exactly what is supposed to improve.
• Source: link the underlying example, not only a retelling.
• Independence: identify which posts share the same evidence.
• Conditions: record query, date and known context.
• Unknowns: separate missing information from observed facts.
• Next test: define what result would change your confidence.

Build pages that deserve to be cited

Google says the established SEO fundamentals apply to its AI features; there are no extra technical requirements specifically for those features. That guidance does not promise inclusion or a first-place ranking.

Our proposed approach is to publish original explanations, clear definitions and inspectable methods, then measure relevant discovery and business outcomes. Keep the query set and observation dates in the evaluation. Treat small samples as exploratory and avoid claiming that one content change caused every movement.

This article is a method for evaluating advice. It reports a repeated-evidence finding, not a successful AI-ranking intervention.