Free Growth Assessment
SEO & AEO

Marketing Attribution for Tech Companies: Crediting SEO, AI Search and Referrals

Analytics hands the credit to the last click, and SEO, AI answers and referrals rarely end there. Here is how to credit each one without guessing.

By Holly Mack Last updated: November 16, 2026 10 min read

Marketing attribution for tech companies is the rule you use to decide which touch earns credit for a deal. Pair analytics, CRM source and what buyers say, and SEO, AI search and referrals stop hiding in Direct.

A woman in a charcoal blazer laying a row of blank cards in a line on a walnut table beside a brass lamp
Credit depends on which touches you lay out and where you decide the line starts.

Here’s what keeps showing up. The monthly report says Direct and Organic Search are the biggest sources of new leads. The sales team says the good ones came from a referral. Both are right.

Nobody’s lying. Two systems are counting different things. One records clicks. The other remembers conversations, and neither one is complete.

On paper, attribution should settle it. Pick a model, read the dashboard, move the budget. Simple enough. In practice the dashboard only sees what got clicked, and the deals worth having in a tech company often start with something that didn’t: a peer’s recommendation, a question typed into an AI assistant, a name that stuck after a conference.

That’s where credit goes missing. And it matters, because whatever gets no credit gets cut. The channel that planted the idea is quietly defunded while the one that closed the sale takes a bow (and yes, this happens more than you’d think).

Disconnected measurement doesn’t compound. Every quarter the budget argument starts from zero.

The rest of this post is a way to stop that. It covers who gets credit for SEO, for AI search and for referrals, and how to check each answer against something a human actually said.

Key Takeaways

The Short Version

Keep three ledgers: what analytics recorded, what the deal record says, and what the buyer told you. Credit SEO by its role in the path, treat AI search as a floor you can’t fully count, and record referrals by name. Compare all three once a month on one page. The reporting side lives in our guide to SEO reporting for tech companies, and the money side is in MSP marketing ROI.

Marketing Attribution Plainly: A Rule for Handing Out Credit

Marketing attribution is the practice of assigning credit for a lead or a deal to the touches along a buyer’s path. Google’s analytics help describes an attribution model as a rule, a set of rules or a data-driven algorithm that decides how that credit gets split.

A rule. Not a fact. Change the rule and the same deal credits a different channel, which is how two reports on one quarter can disagree and both be correct.

Three jobs hide inside the word. Reporting. Planning. And a third that gets skipped, which is settling the argument between marketing and sales. That one decides whether anyone trusts the other two, because a rule that can’t survive the argument between marketing and sales isn’t a system but an opinion with a nice chart.

What Google Analytics Does With the Credit

If your site runs Google Analytics 4, it applies a rule whether you chose one or not. The property’s default is the data-driven model, per Google’s page on the default channel group, and the Attribution reports offer three options in total. Three rules, then.

Where the Credit Lands in GA4

ModelHow it creditsWhere a tech services deal slips
Data-drivenSpreads credit across touches using your own path data, converting and non-convertingIt learns from paths it can see, and a deal that began in a conversation has no path to learn from
Paid and organic last clickGives 100% to the last channel clicked before the key event, ignoring direct visitsThe closing click collects everything, and the referral and the AI answer before it get nothing
Google paid channels last clickGives 100% to the last Google Ads click, and falls back to the rule above when there is noneMatters only if you run Google Ads, otherwise it behaves like the model above

If someone on your team still talks about linear or time-decay credit in GA4, they’re remembering an older tool, because Google’s help page says the first click, linear, time decay and position-based models stopped being available in November 2023. Good to know.

One more rule sits under all three. It’s a quiet one. Every GA4 attribution model leaves direct visits out of the credit unless the whole path was direct, so a buyer who reads two of your posts and later types your address in doesn’t hand the credit to Direct, because it slides back to the last thing they clicked.

Direct still shows up as a channel in your traffic reports, which Google defines as a saved link or a typed URL. Think of it as a bucket for “we can’t tell”, not as a source.

Here’s a made-up deal. There are no numbers, so no one can argue with them. The owner of a managed IT firm hears your name from a peer at a conference. A week later she asks an AI assistant who helps tech companies with SEO and clicks one link, and two weeks after that she searches your name, reads a post and leaves. A month on, she types your address and fills in the form.

What does the report credit? The last click before the form, which is the organic visit. The AI answer gets a slice at best. The peer gets nothing, because a conversation has no click. Your sales rep writes “referral” on the deal, and now two records disagree about one buyer.

At first, this looks like a reporting glitch. It isn’t. Each record is faithfully describing a different slice of the same path.

Crediting SEO Without Over-Crediting It

SEO shows up in two roles, and they deserve different credit. Sometimes a buyer who has never heard of you types a question and lands on your page. That’s SEO starting the path. Other times a buyer who already knows your name searches it to find the site. That’s SEO finishing the job.

Search Console can split the two. Its Performance report groups clicks by query, so you can sort searches that include your company name from searches that don’t. Treat the first group as a closer and the second as a starter. That’s a working view, not a Google rule, but it fixes the usual mistake of handing a branded search the credit for a referral’s work.

Two facts about the categories are worth knowing. GA4’s Organic Search channel includes clicks from Google’s AI Overviews and AI Mode, and Search Console reports that traffic in the Performance report under the Web search type, as Google’s page on AI features explains. So an SEO report can’t separate classic results from AI answers inside Google. Say so.

A simple rule works. Credit SEO in full when a non-branded search started the path. Log a branded search as an assist unless the deal record says search was the first touch. And keep SEO in the conversation about MSP marketing ROI, because a channel that starts paths can look weak in every last-click report and still be the reason the pipeline fills.

Last-click reports reward whoever stands at the door. They say little about who built the road.

Crediting AI Search When the Click Never Comes

Google’s default channel list includes a group called AI Assistant. Google describes it as visits arriving from sources like ChatGPT, Gemini, Deepseek, Copilot or Grok, and it leaves out Google’s own AI Overviews and AI Mode, which stay in Organic Search. Default channel groups can’t be edited, so what you see is what Google classifies.

Good news, then. Some of it is countable. Now the catch. A visit counts there only when analytics can tell where it came from, and a buyer who reads an answer, remembers your name and types it in a day later doesn’t arrive as an AI visit. They arrive as Direct, or as a branded search.

The buyer side shows why that matters. Gartner’s March 2026 release on a survey of 646 B2B buyers, fielded in August and September 2025, reports that 67% prefer a rep-free experience and that 45% used AI during a recent purchase (Gartner). Those are B2B buyers broadly, not MSP buyers, so read the figures as a direction and not as your own numbers. The direction is hard to ignore. Plenty of the research happens before anyone talks to you, and some of it happens inside an answer.

So treat the AI Assistant channel as a floor, never a total. The count tells you AI search is sending visits. It can’t tell you it isn’t sending more. Two other signals fill in the picture. One is how often buyers name an AI assistant when you ask them, which comes up again below. The other is whether searches for your company name are climbing in Search Console, and that one is a signal, not proof, because plenty of things lift branded search.

How you get cited in the first place is a different subject, and how to measure AI search visibility for an MSP covers it. If you wonder why some answers never produce a click at all, that’s the story of zero-click search.

Why Referrals Never Show Up in a Click Report

Same word, two meanings. In GA4, Referral is the channel for visits that arrive through non-ad links on other websites, such as blogs and news sites. In your sales team’s mouth, referral means a person said your name.

One leaves a click. The other leaves a phone call, a text, a LinkedIn message or a warm introduction over email. When the buyer then visits your site, analytics sees an ordinary visit and files it under organic or Direct, because nothing in the visit says a person made the introduction. The referral isn’t missing from the story. It’s missing from the report.

Tech services firms lean on referrals, and referrals have the worst paper trail of any channel. If your PSA closed tickets without recording who opened them, you couldn’t staff the help desk. A deal with no recorded source is the same blind spot. Only more expensive.

Two business owners talking across a small round table with two coffee cups and a closed notebook
The cheapest attribution data is the answer a buyer gives when you ask.

The fix is dull. It works. Make the referrer’s name a required field on the deal record. Ask on the first call, write the answer down in the buyer’s own words, and thank the referrer before anyone else does. A structured MSP referral program turns that habit into a system.

Ask the Buyer: Two Questions That Do Different Jobs

Self-reported attribution sounds like a survey. It’s closer to listening. You add a question to the contact form and to the first-call notes, and you let the buyer name the touch they remember.

Two questions, two jobs. “How did you first hear about us?” tells you where discovery happened. “What made you reach out today?” tells you what tipped them. A podcast episode can be the first answer, and a failed renewal with a current provider can be the second, so you want both and you want to keep them apart.

  • Use a dropdown plus a free-text box. The dropdown keeps the answers countable, and the free text is where the useful surprises turn up.
  • Add an option for an AI assistant, labeled with “ChatGPT, Gemini, Copilot or similar”, and an option for a referral that asks for a name.
  • Let people pick more than one. Memory is lumpy, buyers recall the last touch and the loudest one, and a multi-select box is more honest even though the percentages stop adding to 100.
  • A direction, not a verdict.

Memory is the weak spot. Buyers forget, round off and credit whatever they saw last. Which is exactly why you don’t use this alone. The point is the gap between what the buyer says and what the click says, because that gap is where the hidden channels live.

A One-Page Monthly Reconciliation for Marketing Attribution

Put the three sources side by side and the picture gets honest fast.

The Three Ledgers

LedgerWhat it recordsWhat it answersWhere it’s blind
Click ledgerGA4 and Search Console: visits, channels and queriesWhich pages and channels brought visitsConversations, offline introductions, and AI answers with no click
Deal ledgerThe CRM: the first source, frozen when the deal is created, plus the referrer’s nameWhere deals came from and what they were worthWhatever the rep forgot to log
Conversation ledgerWhat the buyer says on the form and the first callThe touches nobody could trackMemory and wording

Then run it monthly. Our SEO report template gives you the search half of that page.

  1. Tag every link you control. Use utm_source, utm_medium and utm_campaign as Google’s URL builder guide describes, and keep one naming sheet so LinkedIn isn’t spelled four ways.
  2. Find out how AI traffic arrives in your own property. Look for the AI Assistant channel, and note what lands in Direct.
  3. Add the two questions to the form and to the first-call checklist.
  4. Freeze the first source on the deal record when it’s created, and keep a separate field for the latest touch.
  5. Once a month, write one page covering what each ledger says and every deal where they disagree.

Mind the clock, too. GA4 has a key event lookback window, and a deal that takes longer to close than the window loses its early touches, so the deal record has to remember what analytics forgets.

Disagreement between the ledgers is the data. A spike in branded searches with no matching form answers points to a hidden channel. A run of “referral” deals sitting next to Direct traffic points to your introductions. Nothing is broken. That’s the report working. The monthly-report side of this is in our guide to what a monthly SEO report should show.

My working view is to move budget only when two of the three ledgers agree. One ledger can raise a question. It shouldn’t win the argument alone.

What Marketing Attribution Should Change in Your Budget

Three decisions follow. None needs a fancy tool.

  • Don’t cut SEO because it looks quiet in a last-click report. Check whether it starts paths first.
  • Don’t buy AI-search work on the strength of the AI Assistant count alone. That count is a floor. Ask buyers what they say before you pay for the ceiling.
  • Close the loop with referral partners. Telling a partner what happened after their introduction costs very little, and it’s the only way the next introduction gets easier.

More tactics won’t fix a measurement gap. Fixing the gap shows you which tactics were working all along.

So the better question becomes which touch started this, not which one finished it. Less exciting than a new dashboard. More useful, too. If you’d like a second set of eyes on how your current reports credit the work, book a growth strategy call.

What Owners Ask About Crediting Deals

Is there one right attribution model, or do you just pick one?

Short answer: none is right on its own. GA4 offers data-driven, paid and organic last click, and Google paid channels last click, and each is a different rule for the same path.

Stay with the default your property already uses, write down what it leaves out, and lean on the deal record and the buyer’s own words for what a click can’t see. The rule matters less than knowing its blind spots.

Can GA4 actually see traffic from ChatGPT and other AI tools?

Partly. Google’s default channels include an AI Assistant group for visits from sources like ChatGPT, Gemini, Deepseek, Copilot and Grok.

Visits that arrive without a readable source land in Direct or in branded search, so treat the number as a floor, and check how it shows up in your own property before you quote it.

Our best leads say referral and analytics says Direct. Who is right?

Both. About different things. Direct is where analytics files visits it can’t trace, and referral is how your buyer remembers the introduction.

Trust the deal record for where the deal started. Use analytics to see what the buyer did on your site after the introduction.

Should the “how did you hear about us” field be required?

Required, with an “I’m not sure” option so nobody invents an answer just to get past it.

A box that forces a guess pollutes the data it was meant to clean. Add the free-text line, keep it short, and let the first call fill in what the form missed.

Do we need a paid attribution tool to start?

Not to start. A spreadsheet, a required CRM field and the reports you already have in GA4 and Search Console will answer the first round of questions.

Buy software once the monthly page is stable and you can name a question it would answer that the page can’t. Before that, a tool just automates the confusion.

How do you credit a deal that takes six months to close?

Start the clock at the first touch, not at the form fill. Freeze the first source on the deal record, then log the touches that follow.

Analytics lookback windows are finite, so the record you keep yourself is the only one that spans the whole cycle. Review the deal when it closes and ask the buyer what happened in between.

Marketing Attribution SEO Reporting AI Search Referral Tracking
Share
READY TO ACT?

Find out which channel is really getting the credit

Talk to us about how your reports credit SEO, AI search and referrals, so next quarter’s budget follows the deals.

Book a Growth Strategy Call