Skip to main content
Lesson 6 of 8

Lesson 6: Reviews, Mentions and Citations That Feed AI Answers

How off-site signals feed the facts and sentiment AI assistants repeat.

11 min read Free, no signup Written by Adam Yong
Review text, directory listings and editorial mentions flowing into one synthesized answer about a business
4.9 (200+ Verified Business Audits)
100% Free, no signup, no paywall
9,400+ readers
8 lessons · ~80 minutes
What it fixes

Off-site signal problems this lesson addresses

One complaint theme keeps surfacing

A single bad review barely registers. A phrase that repeats across twenty of them becomes an attribute an assistant is willing to state out loud.

You are listed in forty places and half are wrong

Every mismatched citation is a contradiction an engine has to resolve, and the usual resolution is to hedge or pick someone else.

Asking for reviews feels like it might break a rule

Incentives and happy-customer-only requests are both prohibited, and the patterns are recognisable. This lesson stays inside the rules so a filter never costs you the lot.

What this lesson covers

Reviews as a direct AI input

Structured citations versus unstructured mentions

Where it is actually worth being listed

A review request and response cadence

What a negative review really costs

What other people say about you often outweighs what you say

Lessons 2 to 5 covered assets you control. This one covers the rest of your footprint, which is where a surprising amount of the decision gets made. When an assistant is choosing between two similar businesses, independent corroboration is the tiebreaker, because your own site is the one source that is guaranteed to be flattering.

Three kinds of off-site signal matter: reviews, structured citations and unstructured mentions. They work differently and they are worth different amounts of your time.

How AI reads reviews

Not as a star average. Modern systems run aspect-based sentiment analysis, which pulls out the attribute being discussed and the sentiment attached to it. A review that says “arrived on time, explained the fault before touching anything, and the price matched the quote” produces three usable attributes: punctuality, communication, pricing accuracy.

A review that says “great service, highly recommend” produces nothing extractable. It counts towards your average and that is all it does.

This is why a handful of specific reviews beats a pile of generic ones. It also explains the pattern effect: one angry review among forty calm ones is an outlier. Four reviews across six months all mentioning the same missed appointment window is a theme, and a theme is what gets repeated in an answer. The guide on how AI reads your reviews goes deeper on which attribute words recur for local trades.

Review velocity matters in a mild way. A steady trickle looks like an operating business. Twenty reviews in one afternoon after eighteen months of silence looks like a campaign, and platform filters agree.

Asking for reviews without breaking the rules

The rules are short and worth knowing exactly, because the penalties are worse than the shortfall.

You may not gate, meaning you may not screen for happy customers and only ask those. You may not incentivise, including discounts, entries into draws or free add-ons. You may not write them, and you may not have staff write them. You may not bulk-solicit from a purchased list.

What works instead is unglamorous. Ask everyone, at the same moment, in the same way. The best moment is right after the work is finished and the customer has said they are happy. Keep the message to two sentences and a direct link. Then reply to what arrives, including the negative ones, calmly and with specifics.

The guide on asking for reviews compliantly has SMS and email templates you can adapt, plus how to handle a bad review when it lands.

One thing you may do: invite specifics honestly. “If you have a moment, it helps other people if you mention what we worked on and which area you are in” is a legitimate request. Scripting the review is not.

Citations versus mentions

A structured citation is a listing of your name, address and phone on another site: a directory, a trade body, a chamber listing, an aggregator record. Its job is corroboration of your facts, which makes it an extension of the NAP consistency work in Lesson 2.

An unstructured mention is your business name in ordinary text: a local news piece, a supplier’s case study, a community group’s thank-you post, a forum answer. It carries different information, usually about what you did rather than where you are.

Structured citationUnstructured mention
What it corroboratesYour facts: name, address, phoneWhat you do and how you are regarded
How you get itClaim or submit a listingEarn it, or be the person who answers
Where returns stopAfter roughly a dozen quality listingsRarely, because few competitors do it

Both reach an answer, by different routes. The mistake is assuming more citations always means more visibility. It does not. Returns flatten quickly once the major data aggregators such as Data Axle, Neustar Localeze and Foursquare are correct and the obvious platforms are claimed. Past that point, extra low-quality directories mostly create new opportunities for your details to drift out of sync.

The comparison of citations and brand mentions sets out where to spend your next ten hours, which for most businesses is two or three genuine mentions rather than thirty more listings.

Where it is actually worth being listed

Work down this list and stop when it stops being genuine:

  1. The major data aggregators, because smaller directories refresh from them.
  2. The platforms your customers actually use, which may include Yelp for Business, a trade-specific directory, or a booking platform in your sector.
  3. Your local Chamber of Commerce or equivalent, if you are a member.
  4. Trade associations and licensing bodies you genuinely belong to.
  5. Suppliers and manufacturers who list approved installers or stockists.
  6. Better Business Bureau or the local equivalent, where it has real standing in your market.

Then stop. A business with six accurate, relevant listings is in better shape than one with sixty inconsistent ones.

Earning mentions without a PR budget

Mentions are harder to manufacture, which is exactly why they carry weight. Three approaches that work for small local businesses:

Be the source. Local journalists and trade publications need someone who can explain a seasonal problem in plain language. Answering promptly once gets you called again.

Be useful to an organisation. Sponsoring a junior team, running a free clinic session, or fixing something for a community group produces a genuine mention on a site with local standing.

Be listed by people you work with. Suppliers, manufacturers and partner trades often maintain lists and will add you if you ask.

None of this scales, and that is the point.

What to do next

Set the review routine up properly, reply to your backlog, and pick two places where a mention would be genuine. Then move to Lesson 7, which is about telling whether any of the last six lessons actually landed, and what still cannot be measured.

Review text with attribute phrases highlighted and tagged as sentiment aspects
Review text with attribute phrases highlighted and tagged as sentiment aspects
Visual reference

Figures from this lesson

Diagrams you can screenshot and keep beside you while you work through the steps.

Why it matters

Why third-party corroboration carries the weight it does

It is the part you cannot fake

Independent corroboration is exactly what engines are looking for, and that is why it carries weight.

It is bounded work

Chasing hundreds of directories is a waste. Once the aggregators and a handful of relevant sites are right, you stop.

It compounds with the profile work

Every citation that matches your profile makes the whole footprint more consistent, which is the core signal.

It is safe when done properly

This lesson stays inside platform rules, because a review pattern that trips a filter costs more than it gains.

Steps

How to build reviews, citations and mentions safely

  1. 1

    Fix the aggregator records first

    Consistent details upstream stop old data reappearing in the directories that refresh from them.

  2. 2

    Choose five places worth being listed

    Trade bodies, local press, suppliers, community organisations. Relevance beats volume by a wide margin.

  3. 3

    Build the review ask into your workflow

    One trigger point, one short message, no incentives and no filtering. Consistency matters more than volume.

  4. 4

    Reply to everything, especially the bad ones

    A calm, specific reply is readable context an assistant can use, and it changes how a complaint theme reads.

  5. 5

    Re-check twice a year

    Listings drift, platforms change, old records resurface. A short audit twice a year keeps it stable.

Signals in place? Find out whether it worked.

Lesson 7 gives you a 15-minute monthly test instead of a dashboard subscription, and is honest about what cannot be measured.

Reader feedback

What readers said about the off-site lesson

Feedback from owners and SEOs working through the lessons. These are reader comments, not Google Business Profile reviews.

Rated 5 out of 5
"We stopped chasing directory listings and spent the time replying to reviews and getting two genuine local mentions instead. Much better use of an afternoon."
Dana R.
Owner, two-location dental practice
Rated 5 out of 5
"The section on what a negative review actually costs saved a client from a panic. It is a pattern problem, not a single-review problem."
Marcus T.
Freelance SEO, 12 local clients
Common questions

Reviews, citations and mentions: common questions

Does one bad review ruin my AI visibility?

No. A single outlier against a consistent pattern carries little weight. A recurring complaint theme across many reviews carries a lot, because aspect-based sentiment analysis picks up the repeated attribute rather than the star rating.

Can I offer a discount in exchange for a review?

No. Incentivised reviews break Google's rules, and a cluster of reviews that appear after an offer is a recognisable pattern. The downside risk is far larger than the upside.

Is it allowed to ask only my happy customers?

No. That is review gating and it is explicitly prohibited. Ask everyone at the same point in the job, and accept the results.

How many directory listings do I need?

The major data aggregators plus a handful of genuinely relevant local or trade directories. Beyond that, returns flatten sharply and inconsistency risk rises.

Does a mention without a link count?

Yes. Assistants read text, so an unlinked mention on a relevant site still carries information about who you are and what you do. It will not pass link equity, but that is not the mechanism at work here.

Do review replies actually matter?

They add readable context an assistant can use, and a calm, specific reply changes how a complaint reads to anyone assessing the pattern, human or machine.

When is the best moment to ask?

Right after the job is finished and the customer has said they are happy, while the detail is fresh. A review written a month later is vaguer, and vague reviews carry less usable information.

Portrait of Adam Yong

Adam Yong

Author & Founder, LocusPilot

Founder of LocusPilot and Agility Writer; leads GEO strategy at ADE Marketing.

Founder, LocusPilot (AI website builder for local businesses)

More about the author
Open access · Immediate start

Ready for Lesson 7?

Measure Your AI Search Visibility. It picks up exactly where this lesson stops.

No signup. No paywall. 8 lessons, ~80 minutes.