How Perplexity, ChatGPT And Gemini Pick Their Sources

Watch the source list as closely as the mention rate, because it usually moves first. New citations from a directory you corrected are a leading indicator, and they typically appear a month or two before any change in whether you are recommended.

Weeks Three and Four: The Access Findings A technical report covering crawler permissions, what the relevant agents actually receive from your server, whether bot management is interfering, and what survives on your key pages with JavaScript disabled.

And read the raw text periodically rather than only the tallies. Changes in how you are described, from hedged to definite or from generic to specific, often precede changes in whether you appear at all, and no counting method will surface that. Get Recommended by ai

Gemini and Google Surfaces Closest to conventional search infrastructure, which has a practical consequence: work that improves your standing in Google search tends to carry over here more than it does elsewhere.

It also appears more conservative in commercial categories, hedging or declining to make a direct recommendation more often than the others. Where it does recommend, established entity signals seem to matter, which favours brands with consistent details and long records over newer entrants.

One caveat worth writing on the report: any figure produced by a third party visibility tool is a sample from that tool's own prompt set and infrastructure, not a census. Attribute it to the tool by name whenever you quote it, and never present it as a count of what happened. Get Recommended by ai

Most engagements are judged too late, on a final outcome that arrives after the point where anything could have been corrected. The first quarter has its own deliverables, and knowing what they are lets you tell early whether you have hired the right people.

If the baseline exists, access problems were found and fixed, listings were corrected with names attached, and the source list has begun to move, the engagement is on track even if mention rate has not shifted. If none of those happened, the next ninety days will not be different from the first. Get Recommended by ai

One scheduling detail improves comparability more than it should. Run on roughly the same date each month rather than whenever somebody remembers. Retrieval behaviour and the freshness of competing sources both vary over a month, and a series taken at irregular intervals introduces variation that looks like a trend.

Log the conditions with every run, including which assistant, which mode, whether web access was enabled and the date. When a result moves sharply, the conditions log is usually what tells you whether the world changed or your setup did.

Why One Snapshot Proves Almost Nothing Generation involves randomness, and retrieval can return different pages between runs. The same prompt asked twice in a row can produce different companies in different orders.

Run each prompt at least three times. Assistants vary their answers between runs, and a single result is a sample rather than a finding. Record the full text of each answer and every source cited, not a summary.

Making Any Model Safe Four clauses do most of the protective work regardless of structure. The prompt set and baseline archive belong to you and leave with you. Raw answers ship with every report. Scope is stated in countable units. And there is a defined review point with agreed criteria before the contract auto renews.

This is a working method you can run yourself in an afternoon, repeat monthly, and hand to an agency as a brief. It produces a record you can argue with, which is more than most reporting in this field manages. Get Recommended by ai

The condition is that the output has to be yours to keep and act on elsewhere, including the prompt set. An audit that only makes sense inside that agency's retainer is a sales document with a price attached.

How to Run the Ninety Day Review Ask three questions. Can you show me the prompt set is unchanged. Can you show me the raw answers. What specifically did you do, and which of the changes do you believe caused which movement.

The honest framing first: nobody outside these organisations knows the selection logic, and the systems change without announcement. What follows is drawn from observable behaviour, visible citations and published research, which supports useful generalisations and does not support precision.

The specific damage is that somebody sees a dip, rewrites a page, sees the number recover for unrelated reasons, and concludes the rewrite worked. That false lesson then gets applied elsewhere. A slower cadence with more runs per prompt is more informative than a faster one with fewer.

The Objection, and the Answer to It Sales teams resist naming competitors and conceding anything, and the resistance is understandable. The counter is that the comparison is happening regardless, inside a model, using whichever sources it can find.

Set a Cadence and Stick to It Monthly is enough for most categories. Run the same prompts, the same number of times, and keep every answer. The value compounds because you can look back and see when a competitor entered the shortlist and which source appeared alongside them.