Guides

What Is a Social Search Agent? How It Differs From a Search Engine

A social search agent takes a question, plans and runs several searches, reads the results and answers with citations. How it differs from a search engine.

A social search agent is a chat that takes a question in plain words, plans and runs several social media searches, reads the results and answers with citations to the posts it found. A social search engine does one thing per request: it returns results for one query on one network. The difference is who does the planning and the reading. With an engine it is you; with an agent it is the software, inside limits you set.

Search engine vs search agent

A social search engine fits a lookup: you pick the network, write the query and read the list. An agent sits on top of the same searches: it decides which to run, runs them and writes an answer that points back to what it read.

What changes when an agent does the searching
Search engineSearch agent
You give itOne query: a keyword, #hashtag, @handle or linkA question in plain words
What runsOne search on one networkSeveral searches, chosen for the question
Who reads the resultsYouThe agent first, then you through its citations
What you getA list of posts, accounts or commentsA written answer with numbered citations
Cost controlThe exact price shown before the search runsA credit budget per chat, and a price ceiling on every plan card
FitsA lookup you can phrase as a queryAn open question that needs several searches

How Superlurk's agent works

In Superlurk the agent is in the app's Agent section: a chat that looks things up as you talk. You ask in plain words; it runs the searches, reads what comes back and answers with the posts it found. A chat goes like this.

  1. Set the budget. Each chat has a credit budget that you set. The default is 25 credits, and the agent works inside it.
  2. Ask the question. Write it the way you would ask a colleague, for example “what do people complain about with standing desks?”
  3. Let it run the 1-credit searches. The agent runs 1-credit searches on its own inside the budget: up to 4 per Quick reply and up to 8 per Deep reply.
  4. Approve anything bigger. A more expensive search, or any multi-step job, comes back as a plan card marked “Plan · nothing spent yet”. It lists the steps, shows “Up to N credits” and waits for you to press Run.
  5. Read the answer and open the citations. Every claim carries a numbered citation. Opening one shows the post or comment the claim rests on.

Quick and Deep replies

The two reply modes
QuickDeep
Price per reply1 credit5 credits
1-credit searches it can run on its ownUp to 4Up to 8
AnswerShort, from a few searchesFuller, from more searches
Use it whenThe question is narrowThe question has several parts or spans networks

A reply with no supported finding and no plan is free. Agent replies use the same credits as searches and workflows, so there is one balance to watch. Plans are on the pricing page.

How to check an answer

The agent links every claim to the post or comment it rests on, through numbered citations that open the cited item. That is also how you check an answer.

  • Open the citations behind the claims you plan to repeat, and read the whole post, not only the quoted line.
  • Check the date and the author of each cited item.
  • Count how many items a claim rests on. Two posts are an example, not a pattern.
  • Ask a follow-up when support looks thin; the agent can run another search inside the budget.

When a plain search is still the better tool

  • You already know the network and the query: a keyword, a hashtag, a handle or a link.
  • You want the list of results itself, saved to your private Results and exportable as CSV or JSON.
  • You want to pay for exactly one search. Most cost 1 credit, the exact price is shown before anything runs and a failed search is free.
  • You will repeat the same search on a schedule, which is a job for a monitor.

What an agent cannot do

  • Public content only. It cannot see private accounts, deleted posts or anything a network does not make public.
  • Counts are first-page counts. A figure in an answer comes from what the first page of each search returned, not from the total volume of posts on a network.
  • It answers when you ask. It does not run on a schedule, and new-post alerts are not available.
  • Chats are kept for 30 days. Note the posts you need before then.

From an answer to a repeatable report

When a plan worked and you expect to need it again, save it as a workflow: the same steps, ready for the next brief. Superlurk also has ready-made research workflows for common reports, explained in the workflow guide. If you are unsure which tool fits a job, see search, agent, workflow or monitor.

Frequently asked questions

What is a social search agent?

A social search agent is a chat that runs social media searches for you. You ask a question in plain words; it plans the searches, runs them, reads what comes back and answers with links to the posts and comments it used.

How much does the agent cost in Superlurk?

A Quick reply costs 1 credit and a Deep reply costs 5 credits. The agent runs 1-credit searches on its own inside a credit budget you set for each chat (25 credits by default), and a reply with no supported finding and no plan is free.

Can a social search agent see private accounts?

No. The agent searches public content only, the same as a plain search. It cannot see private accounts, deleted posts or anything a network does not make public.

How this guide was written. Written by the Superlurk team on 12 November 2026 from the app's current behaviour. Examples are illustrative, not measured customer results.

The Superlurk TeamSocial research

We build Superlurk for public-data research across 25 supported social networks, with the price shown before every search. We write practical guides to search methods, source verification and AI-assisted research.

Start lurking for free.

Every network, no card needed. See the price first, and pay nothing if a search fails.