Social listening is the job of finding who is mentioning your business, your competitors and your category, then acting on the few mentions that matter. It takes a Tech and Skill shape: an afternoon to set up free monitors, then about half an hour a week where the AI you already pay for reads what they caught and flags what needs you. The monitors do the collecting, because a chat assistant is not built to watch groups, threads and review sites between conversations. The AI does the reading, because sorting a week of mentions by hand is the part that gets skipped. What stays with you is the reply, the decision to post it, and anything that touches a real customer.
- Shape
- TechSkill
- Sitting
- an afternoon
The shape: Tech and Skill
This job has two halves, and each one takes a different shape.
The collecting half is Tech. Something has to watch the web, Reddit, local groups and review sites, and that something is a set of free alerts, review notifications and saved searches. You build them once. After that, the alerts and review notifications come to your inbox on their own, and the saved searches wait as bookmarks you open once a week.
The reading half is Skill. A pile of raw mentions is mostly noise: a different business with a similar name, a news article that uses your category word in passing, a spam post. Reading that pile and pulling out the three items that matter is judgment work, and it is the work the AI you already pay for does well when you give it clear rules.
What this means for you: spend this afternoon on the Tech half, and every week after that the job shrinks to one paste and one read. If you skip the setup and ask a chat assistant to search for mentions live, you will get a short list of whatever a web search happens to return, and it will miss most of what people say in groups and threads.
Setting up the monitors
Start with free tools that deliver on a schedule.
- Google Alerts watches the open web for a phrase and emails you as it happens, once a day or once a week. Set one per phrase, set the frequency to weekly, and choose "All results" rather than "Only the best results".
- Reddit search sorted by new, saved as a bookmark for each phrase, shows recent posts and comments in local and industry communities.
- The search bar inside each platform where your customers talk: Facebook groups you belong to, LinkedIn post search, Instagram and TikTok search for your name and your category plus your city.
- Your review profiles: Google Business Profile, Yelp, and any industry directory that lists you. Turn on review notifications where the platform offers them.
Keep the phrase list short. Your business name, its common misspellings, your own name if customers know you by it, two or three competitors, and two buyer phrases, such as "recommend a bookkeeper in Tempe" or "anyone know a good mobile groomer".
Then make one plain document called "Mentions" and, once a week, paste into it whatever the alerts and saved searches turned up. Copy the text, the platform, the date and the link for each item. That document is the input for the Skill half.
What you give the AI
Give it these five things, by name, in one message.
- This week's mentions document: every item the monitors caught, with the text, the platform, the date and the link. Raw is fine. Do not pre-sort it, because sorting is the job.
- Your phrase list: the exact names and phrases you track, so the AI can tell a real mention of your business from a stranger with a similar name.
- A two-line description of your business: what you sell, where, and who buys it. This is what lets the AI recognize a buyer question in your category.
- Your competitor list: the two or three names you watch, with one line each on how they differ from you.
- Your sorting rules: what counts as urgent for you (a public complaint, a safety issue, a reporter), what counts as an opportunity (someone asking for a recommendation in your area), and what you never want to see again (job postings, a business in another state with your name).
Leave out private customer records, account numbers and anything about a customer that is not already public. The AI only needs what the public saw.
What comes back
You get back a short triage list in four groups, with a link and one line of reasoning for each item, plus draft replies for the items you might answer.
Paste this brief, then paste your mentions document under it.
You are reading one week of public mentions for a small business.
Your job is to sort them so the owner reads only what needs them.
The business: [two lines: what we sell, where, who buys it].
Names and phrases we track: [your phrase list].
Competitors we watch: [names, one line each].
Urgent means: [your rules, for example a public complaint, a safety issue, a reporter asking].
Opportunity means: [your rules, for example someone in our area asking for a recommendation].
Ignore: [your rules, for example job posts, a business with our name in another state].
Sort every item into exactly one group:
1. Needs a reply this week
2. Opportunity worth a reply
3. Worth knowing (competitor news, a question people keep asking, praise we could reuse with permission)
4. Ignore
For every item in groups 1 to 3, give the link, the platform, the date, and one sentence on why it is in that group.
For groups 1 and 2, draft a reply of three sentences or fewer, in plain words, that answers what the person asked before it mentions us.
If a reply would mention our business, say so openly ("we run ...").
For group 4, give only a count and the most common reason.
Then list any question that appeared more than once, in the asker's own words.
Do not invent facts about our business, our prices or our customers.
If an item is unclear, put it in group 3 and say what is unclear.
Keep group 3 to five items or fewer; move the rest to group 4.The return is a list you can read in five minutes. Most weeks, group 1 is empty or has one item, group 2 has one or two, and group 4 holds most of the pile. The repeated questions at the end are often the most useful part, because they are what your next posts and your website FAQ should answer.
An editorial example
This is an invented example, not a customer. Say you run a two-person mobile dog grooming van in Mesa. Your monitors catch a Google Alert for a pet-store blog that names you in a list, a Reddit thread in a local community asking who grooms anxious dogs at home, a Facebook group post praising a competitor's new van, and a dozen alerts for a grooming salon with a similar name two states away. The AI puts the Reddit thread in group 2 with a draft reply that answers the question about anxious dogs first, the blog in group 3, the competitor post in group 3 with a note that people liked the van's photos, and the out-of-state salon in group 4. You read four lines, check the draft against how you actually handle nervous dogs, and decide whether to post it.
How to judge it
A good triage is short, specific and links every claim to a source item. You can reject a bad one in a minute by looking for these signs.
- Group 1 or 2 holds items that are not about you. The AI matched a similar name or a generic category word. Tighten the phrase list or the ignore rules and run it again.
- An item has no link, or a link you did not paste. The AI made something up. Throw out that item and check the rest against your document.
- The draft reply leads with your business. A reply that opens with your name and a pitch reads as spam in most groups. It should answer the person's question first.
- The draft reply states a fact you did not give it, such as a price, a hold time, a guarantee or a service area. Strike it, because you are the only source for those.
- Everything is "worth knowing". A triage that puts thirty items in group 3 has not sorted anything. Ask it to keep group 3 to five items and move the rest to group 4.
- The repeated-questions list paraphrases instead of quoting. You want the asker's own words, because those are the words a buyer types into a search bar.
If two weeks in a row come back mostly in group 4, the problem is the monitors, not the reading. Cut the noisy phrase and add a sharper one.
What stays with you
The AI reads and drafts. These parts are yours, and no AI does them for you.
- The facts. Only you know whether a complaint matches a real job, whether the person in the thread is already a customer, and whether a claim about your business is true. Check every fact in a draft against what you know before it goes anywhere.
- The decision to reply at all. Some mentions are better left alone. A review that touches privacy, a safety issue or a legal question goes to the right person before anyone posts anything. Our guide to responding to negative Google reviews walks through that decision.
- The post button. Paste the reply yourself, from your own account, after you have read it. Follow each group's rules on self-promotion, and say plainly that you run the business when you mention it.
- Permission to reuse praise. A kind post about you is not yours to republish. Ask the person first, then use their words, as covered in turning customer reviews into social proof.
- What you do with the patterns. The AI can tell you that three people asked the same question this month. Deciding to answer it on your website, in a post, or by changing how you work is a business decision.
Free monitors or a paid listening tool
There is a real choice here, and it turns on how many mentions you get.
| Your situation | What fits | Who does the reading |
|---|---|---|
| Fewer than a few dozen mentions a week, one location | Free alerts and saved searches, plus the weekly AI read on this page | You, with the AI sorting |
| Several locations or a well-known name, mentions every day | A paid listening tool with broader coverage and history, price on vendor site | You or a staff member, with the tool's dashboards |
| You want listening, scheduling and reporting in one suite | A larger social suite; see our comparison of Sprout Social alternatives | A staff member or an agency |
Most small businesses sit in the first row. Start there, and move up only when the weekly paste takes longer than the read.
Make it a habit
Put a fixed half hour on your calendar, the same day each week. Paste the week's mentions, run the brief, read the four groups, and act on group 1 before you close the tab. Keep the brief in a saved note so you never rewrite it. Every month, look at the repeated questions and add one of them to your website or your next post. That is the step that turns listening into marketing.
Of every job on this list, social media is the one that comes back every week, and Boomp is software that does that one for you. See ten posts made from your website, free
