You Know Who Visited. Here's What to Do Next.
Website visitor identification tells you which companies, and sometimes which people, read your site without filling in a form. This playbook covers how it works and how to choose a tool, then what to do with each match: a conversation while they're there, a value-first note after they leave, retargeting from what they read, and the account owner brought in at named moments.
In this article
The tool says a target account read your pricing page yesterday. The rep sees the alert after lunch, sends “I saw you were on our site”, and the buyer has already talked to the vendor who answered.
Identification gives you a signal, not a conversation. Match the response to what you know: a conversation while they're there, a value-first note after they leave, retargeting from what they read, and the owner brought in at named moments.
Website visitor identification matches a visit from someone who never filled in a form to a company, sometimes a person, with the pages they read. Pin this ladder above your alert channel and sort each visit by what you know:
| What you know | Do this | Never this |
|---|---|---|
| Target account, on the site now | Start a conversation; the owner takes over at named moments | Name their company before they do |
| Contact identified, left | A value-first note within the day; one question; a person on request | ”I saw you were on our site” |
| Company identified, left | Account-level value, such as the case study for their industry; alert the owner | A demo ask in message one |
| Anonymous, left | Retarget by page group; a conversation that knows the topic when they return | Follow everyone for 30 days |
Can a website tell who is visiting it?
Partly. Tools can often match a visit to a company and sometimes to a person, but never every visitor, and a person-level match is a probability, not a fact. Consent rules decide what you may collect. That partial view is still worth having, because most interest never raises a hand:
- 3.5%of website visitors fill out a form.1
- 61%of the way through their buying process is when buyers make first contact, down from 69% the year before.2
- 80%of deals go to the vendor buyers preferred before engaging sellers.2
So an identified visit is often a shortlist check: confirm fit, don’t pitch from zero.
How do you identify website visitors?
A small script on your site sends each visit to a vendor, which matches it against company network ranges, an identity graph, or contacts you already know.
- Network address to company.The backbone of company matching. It struggles with home networks, phones and VPNs.
- Identity graphs.Where person-level matches come from, and where consent rules bite hardest.
- Known visitors.Someone who clicked your email or filled in a form before. The most reliable person-level signal.
Website visitor tracking in analytics counts visits; identification puts a company name on some of them.
What is the best website visitor identification software?
There’s no single best. Run two tools on your own traffic for a month and judge the matches you could act on: company or person level, match quality on evaluation pages, CRM write-back that respects ownership, consent handling, suppression, and what happens after a match. That last one is where most programs stall. Only half the companies buying intent data used it to accelerate pipeline.3 A list of logos someone checks every few days is the status quo to replace.
What should I do after I identify a website visitor?
Decide which visits matter, agree owners, start a conversation while they’re there, follow up by what they read, and bring the owner in with the full record. Never mention their browsing. The difference, in one alert:
What usually happens
An alert, "Acme Corp viewed /pricing", sits in a team channel until after lunch. Then: "Hi Priya, I saw you were on our site. Do you have 30 minutes for a demo?"
What to do instead
While she's on the page, a greeting that says it's an AI and offers help with rollout or a person. She says who she is. If she leaves, a note within the day about the topic, with one useful thing and one question.
Step 1: Decide which identified visits should trigger action
Write one rule: which accounts, which pages, how recent. Marketing ops writes it with RevOps from the target list and ownership in your CRM. If it fires hundreds of times a day, it’s too loose.
The trigger rule
- Accounts
- on the target list, or with an open opportunity
- Pages
- pricing, comparison, case studies, product
- Recency
- on the site now, or in the last day
- Excluded
- customers, competitors, your own staff, careers and support pages
Read a visit as an account signal: relevance aimed at one individual rather than the buying group has a 59% negative impact on group consensus.4 The failure is chasing every blog reader, or treating a match as certain.
Step 2: Agree who owns each account before anyone reaches out
Every target account needs an owner field routing can read, a round robin where there’s none, and a named list where the owner wants every conversation. A stale owner field sends the buyer to someone who doesn’t know them.
Step 3: How do I use AI to start a conversation with a visitor without being creepy?
The greeting says it’s an AI, offers help with the page’s topic, and offers a person. It never names the company or the browsing before the buyer confirms who they are: more ways forward than a form, without feeling watched.
Disclosure isn’t optional: 87% of customers say a route to a human is essential when a company uses generative AI,5 and from 2 August 2026 EU law requires telling people they’re talking to an AI.6 Before booking, confirm four things in the buyer’s words; what she says outranks anything inferred.
- The greeting names nothing
Agent: Hi, I'm Brightline's AI assistant. Happy to answer questions about rollout, or I can connect you with someone on our team.
Buyer: We need to replace our website chat before Q1. Our vendor is shutting down.
- Four mandatories, in her words
Name and role: Priya, demand gen at Acme. What prompted the visit: replacing a chat vendor. Timing: before Q1. Who else is involved: "Our RevOps lead."
- The owner, not the next free seller
Agent: Jordan looks after Acme. Would 30 minutes on Thursday work? I'll pass Jordan your timeline so you won't have to repeat it.
Illustrative; Acme and Brightline are stand-in names. The check for every line: would the buyer be surprised we knew it? Then cut it.
Step 4: Route target accounts to the owner and name when a person takes over
Book with the account owner, and write down the moments the AI stops; each escalation pauses the AI and alerts the owner with the whole conversation. Sales and marketing sign this table together (for everyone else, see how to increase website conversion rate):
| Moment | The AI | The account owner |
|---|---|---|
| Notices the visit | Starts at any hour | Alerted for named accounts |
| Opens the conversation | Says it’s AI; offers a person | — |
| Confirms the mandatories | Asks, in the buyer’s words | Reads them first |
| Answers questions | From approved knowledge and published pricing | Anything beyond |
| Discounts, terms, open deals | Hands over | Owns it |
| Books and prepares | Books with the owner | Runs the meeting |
Escalate when: an open opportunity · the owner’s named list · the buyer asks for a person · discounts or contract terms.
Step 5: Win back website visitors who leave without converting
Sort the exit with the ladder above, then send something useful at a receptive moment: for a contact you can reach, a note within the day about the topic they read.
Speed still counts. Reaching an account within an hour of it landing made a meeting six times likelier,7 and an older study of six companies put the odds of reaching a lead 100 times higher at five minutes than at thirty.8 But fast and generic does harm: 73% of B2B buyers actively avoid suppliers who send irrelevant outreach.4
Use when: A contact you can reach read a topic page and left. Consent checked for your region.
Subject: {Topic}: the short version
Hi {name}, if you’re weighing up {topic}, here’s {one useful asset} that covers the questions most teams start with. One question: is {the decision} on a deadline? If it’s easier to talk it through, {owner} can take 15 minutes.
Quotes the buyer: the topic, never the visit. No demo ask in message one, and no “I saw you on our site.”
Check your region’s consent rules with legal before writing to someone who never gave you their details; GDPR, ePrivacy and CAN-SPAM differ. If they gave an email or number, continue on the channel they last answered, with memory, and stop on a reply, an opt-out or a booking: three to five useful touches at most (lead follow-up). Marketing writes; sales routes.
Step 6: Retarget website visitors by what they read, not "all visitors"
Build audiences by page group and depth, and run ads that continue the topic they read, landing on a page whose conversation already knows it (landing page personalization). Paid media owns this; third-party connectors such as Facebook’s MCP can carry browsing signals into the ad platform. Set a frequency cap and an end date. The check: one message per audience, tied to its page group.
Step 7: Hand the owner the record and protect the meeting
Send a confirmation with the reason for the meeting in the buyer’s words, and make sure the owner reads the conversation first. Note everyone the buyer named; that’s the start of the buying committee. Reminders and easy rescheduling keep the meeting alive (meeting reminder emails). The buyer shouldn’t have to say anything twice.
How do you measure what happens after identification?
Count each step against the one before it, never the match count alone. The AI system reports counts; rates come from your CRM.
| Metric | Denominator | Period | Source |
|---|---|---|---|
| Conversations started | Identified target-account visits | Weekly | Conversation system |
| Replies or return visits | Value-first notes sent | Monthly | Sending system, CRM |
| Meetings booked | Conversations with target accounts | Monthly | CRM |
| Meetings held | Meetings booked | Monthly | CRM |
| Return visits | Retargeting audience size | Monthly | Ad platform |
| Pipeline created | Meetings held | Quarterly | CRM, no attribution claimed |
Where should you start this week?
- Write the trigger rule for open opportunities only.Two page groups, pricing and comparison, and the exclusion list.
- Clean the owner field on those accounts.Then agree the four escalation moments with sales.
- Put one value-first note in front of legal.Topic, one asset, one question, a person. Steps 3 to 5 come next; retargeting by page group comes last.
How Synapsa turns identified visits into conversations
Everything above works with any capable stack. In Synapsa, the visit and the conversation live in one system.
- It starts from the signal. A site script identifies visitors and their page, never every visitor with certainty, and Conversion OS starts the conversation at any hour. Other identification tools connect through a configured API.
- It confirms before it assumes. What the buyer says outranks enriched data, and the agent checks its mandatories before booking.
- It routes on your CRM. Escalation pauses the AI at once and alerts the people you name, with the full record.
- It follows the buyer by email or SMS with one memory across channels, plus reminders and rescheduling.
- Marketing OS builds the pages and ad content; retargeting runs in your ad tools.
Momentum, a Synapsa customer, reported website conversion moving from 0.2% to 2 to 3%, with AI qualifying visitors before booking.9 That’s one customer’s reported result, not a forecast, and it isn’t an identification result. Across customers: 500+ agents deployed10 and 1M+ buyer interactions.10
Frequently asked questions
What is website visitor identification?
It's matching a visit from someone who never filled in a form to a company, and sometimes to a person, using the visitor's network address, device and browser data, and identity data a vendor already holds. The output is a named account, the pages it read, and sometimes a likely contact. Analytics tells you how many people visited; identification tells you which companies were among them.
Can visitor identification tell me which person visited, or only the company?
Most tools identify the company first. Some add a likely contact by combining the company match with job titles, pages viewed and identity data they hold. Treat a person-level match as a probability, not a fact, and let the buyer confirm who they are in the conversation.
Is website visitor identification legal?
It depends on where your visitors are and what the tool collects. Matching a visit to a company from network data is treated differently from identifying a named person. European and UK rules generally require a lawful basis and clear disclosure; several US states regulate the sale and sharing of personal data and honor Global Privacy Control signals. Disclose the tool in your privacy policy, run it behind your consent banner where consent is required, and involve counsel before writing to anyone who never gave you their details. This article is not legal advice.
How is visitor identification different from website analytics?
Analytics tells you how many people visited and what they did. Visitor identification tells you which companies, and sometimes which people, were among them. One measures the audience; the other gives your team an account to act on.
Sources
- 6sense, 2022 research, reported by Demand Gen Report. Commonly rounded to 3%.
- 6sense, Buyer Experience Reports 2024 and 2025. Nearly 4,000 buyer responses in 2025; vendor research.
- Forrester, Q1 2023 Global B2B Intent Data Survey.
- Gartner, 2025 sales surveys of 632 B2B buyers, fielded August to September 2024.
- Gartner, 2026 survey of 3,566 B2B and B2C customers, service context.
- EU AI Act, Article 50(1), Regulation (EU) 2024/1689, applying from 2 August 2026.
- Drift platform data, 2023. No sample or period stated.
- InsideSales, Lead Response Management study, 2007. Six companies; contact odds, not conversion.
- Momentum, customer-reported result, Synapsa case study.
- Synapsa platform data, 2026.
