Local Businesses And The AI Recommendation Problem

A simple system beats a campaign. Ask every satisfied customer, at the point where they have just been satisfied rather than a month later. Make it one click. Respond to everything, briefly and without defensiveness.

Record the conditions alongside the results: which assistant, which model version if visible, whether web access was on, the date and the run number. When a result changes sharply, the conditions log is usually what tells you whether the world changed or your setup did.

Review the whole set annually rather than continuously. Markets shift, product lines change and language moves, but an instrument revised every month is not an instrument. It is a series of unrelated measurements that happen to share a spreadsheet. llm seo

Consistency Matters More Than Anywhere Else Local identity resolution depends on the business details agreeing across a long tail of directories, many of which nobody has looked at in years. Old addresses, disconnected numbers and previous trading names sit in these places indefinitely.

There is also a straightforward test that costs nothing and tends to end the debate internally. Ask an assistant the question your best customer would have asked before they found you, and read the answer out in the next management meeting. llm seo

Move next to sources that accept corrections, which costs an email each and has a better acceptance rate than most people believe. Only then invest in earning genuinely new coverage, which is the expensive part and should be aimed at the specific publications your baseline showed are already being cited.

Why Local Is More Exposed The classic local query is a recommendation request with a geographic constraint, and that maps directly onto what a generated answer does well. Somebody asking who to call for a specific job in a specific town receives two or three names rather than a map and a list to work through.

The practical result is that a claim appearing only on your website is treated as a claim, while the same claim appearing in a trade publication, a review platform and a forum thread starts being treated as a fact about the world.

There is a variant of this worth checking separately. Sometimes you appear and the competitor appears above you, which is a different problem from being absent. In that case compare the specificity of the two descriptions rather than the sources: the company described in concrete terms tends to be listed first, because a specific description is easier to justify than a general one.

Where to Get Real Language Four sources, all of which you already own. Sales call notes, where prospects describe their problem before anyone corrects their terminology. Support tickets, where customers describe things going wrong in their own words.

There is also a mechanical problem. Manufactured mentions tend to be uniform in language and timing, which is exactly the pattern that gets discounted. The effort produces a body of sources that agree suspiciously well and carry less weight than a smaller number of genuine ones.

Weight toward the commercial tiers. Roughly a third on buying intent, a quarter on evaluation, a quarter on problem framing and the remainder split between definitional and branded is a reasonable starting distribution.

None of these are expensive to fix and all of them are total. A blocked crawler does not reduce your visibility, it eliminates it, and every day the block stands is a day of answers composed without you in them.

What analytics cannot tell you is how often you were named without a click, which in this channel is most of the time. A recommendation that a buyer acts on three weeks later leaves no trace in any report you own. This is why the manual prompt set is not optional, and why nobody should be asked to justify this work on referral traffic alone.

Tracking this is genuinely awkward, and pretending otherwise is how most reporting in this field goes wrong. There is no console. Answers vary between runs. Referral attribution is inconsistent between assistants. Anyone handing you a single confident number has hidden a great deal of variance behind it.

The Baseline Is Worth More the Earlier You Take It A baseline taken today lets you attribute change later. Without one, when something moves you will be reduced to guessing whether it was the assistants, a search update, a competitor's campaign, seasonality or your own site changes.

Being missing from the five pages that generate your category's answers is a complete explanation on its own, and it is fixable without anyone's permission on the platforms that accept claims and corrections.

This is the whole argument in one sentence, and it is why the audit is worth running even if you intend to do nothing with the findings for six months. The measurement is cheap. Reconstructing a baseline you never took is impossible.

Run Each Prompt Multiple Times Generation involves randomness and retrieval can return different pages between runs, so a single answer is a sample. Three runs per prompt is the practical minimum and five is better where the stakes are high.