SEO attribution: why your analytics are lying to you.

A hotel owner sits down with a marketing report. Organic search traffic is flat. "Direct" traffic is up substantially — a big, encouraging number, apparently arriving from nowhere. The conclusion practically writes itself: the SEO investment isn't moving the needle, while the brand is somehow getting stronger on its own. So the content budget gets trimmed, the SEO retainer gets questioned, and the spend shifts toward channels with cleaner-looking numbers. Six months later, bookings soften in a way nobody can quite explain, and the "direct" traffic that looked so healthy quietly deflates along with the content that was actually producing it.

This is not a hypothetical. It is one of the most common and most expensive mistakes in hotel marketing right now, and it happens because of a technical failure almost nobody in hospitality has been told about: a large and growing share of the traffic your content earns does not arrive labeled as such. It shows up as "direct," or "unassigned," or nowhere at all. The channel doing the work gets none of the credit, and the channel getting the credit is doing none of the work. Decisions get made on those numbers. And the predictable result is that hotels defund exactly the thing that is working, because their analytics told them a story that wasn't true.

This is the operator's guide to seeing clearly. Why attribution is breaking in hospitality specifically, where your traffic is actually disappearing, the particular disaster of the booking-engine handoff, how AI-driven traffic hides in your reports, what you can realistically fix, and how to build a measurement picture honest enough to make decisions on. This isn't about which metrics to track in general — for that, see our look at what hospitality marketers actually track. This is about something narrower and more urgent: why the numbers you already have are lying to you, and what to do about it.

What attribution is supposed to do, and why it's failing.

Attribution is the attempt to answer a simple question: what caused this booking? A guest reads a blog post about your neighborhood, leaves, thinks about it for two weeks, asks an AI assistant for a recommendation, searches your brand name, clicks a paid ad, visits on their phone, comes back on a laptop, and finally books. Which of those deserves the credit? Analytics platforms make a rules-based guess — typically crediting the last non-direct source, or apportioning across the path — and present the result with a confidence the underlying data does not deserve.

That guessing has always been imperfect. What has changed is that the raw signal itself is degrading, and degrading fast. Attribution depends on referrer information and on tracking that persists across a visitor's journey, and both are being eroded simultaneously: browsers restrict cross-site tracking by default, privacy tooling strips identifiers, users decline consent, and — crucially for hotels — the new AI-mediated discovery layer often sends visitors with no usable referrer at all.

The result is a growing bucket of traffic your analytics cannot classify. And here is the crucial thing to understand about that bucket: unattributed traffic doesn't get labeled "unknown." It mostly gets labeled "direct." Direct is the default dumping ground — the category for a visitor who appears with no discernible source, as though they simply typed your URL from memory. Some of them did. Most of them, increasingly, did not. They came from somewhere, and that somewhere is now invisible.

Where your traffic actually goes to hide.

The "dark traffic" problem — visits with no reliable source data — has several distinct causes, and it helps to name them, because they have different fixes.

AI assistants and chat interfaces. When a traveler asks an AI assistant for hotel recommendations, reads about your property, and clicks through, the referrer information that arrives is often absent, inconsistent, or unrecognized by your analytics' channel groupings. So a visit that was genuinely earned by your content — cited by an AI because you did the work to be citable — arrives with no fingerprints and lands in direct. This is the fastest-growing source of misattribution in travel, and it is particularly cruel: the more successful your AI-search work becomes, the more your reports will insist that your content is doing nothing and your brand is mysteriously strengthening.

The zero-click answer that doesn't click at all. Sometimes the AI or the search result answers the traveler's question without a visit — your check-in policy, your amenities, your recommendation — and no traffic arrives whatsoever. That visibility is real and valuable and appears in your analytics not at all. It may still lead to a booking days later that arrives as "direct."

Cross-device journeys. A traveler discovers you on a phone during a commute and books on a laptop a week later. The two sessions are frequently not connected, so the discovery gets no credit and the booking looks like it materialized from nowhere. Travel has unusually long, multi-device consideration windows, which makes hospitality more exposed to this than most industries.

Privacy protections and consent. Browsers blocking trackers, users declining consent, ad blockers stripping parameters — all of it removes the signal that attribution depends on. This is not a bug being fixed; it is the direction of travel, and it will not reverse.

Long consideration windows. Travel decisions unfold over weeks. Tracking windows expire, cookies get cleared, sessions break. The blog post that first put you in the consideration set six weeks ago has long since been forgotten by your analytics by the time the booking lands.

Untagged campaigns and messaging apps. Links shared in messages, emails, and apps frequently arrive stripped of referrer data. Your own campaigns, if untagged, come back to you as direct — self-inflicted and completely avoidable.

The visibility that produces no traffic at all.

Before going further, it is worth confronting the hardest case, because it breaks the model most marketers carry in their heads: visibility that generates zero traffic and is nonetheless valuable.

An AI assistant tells a traveler your check-in time, your pet policy, and that your hotel is a good choice for a family staying near the waterfront. The traveler never visits your website. Three days later they book — perhaps directly, perhaps through an OTA. Nothing in your analytics records the recommendation that caused it. The most consequential marketing event in that guest's journey left no trace on any system you own.

The instinct is to call this a loss, but that framing is wrong for a hotel, and it matters that you get it right. A publisher monetizes page views, so a zero-click answer genuinely steals their revenue. A hotel monetizes rooms. Being named, accurately and favorably, in the answer a traveler receives is a win — arguably a bigger one than a click, because a recommendation carries more weight than a link. The problem is not that the visibility is worthless. The problem is that it is invisible to your instruments, which means it will be undervalued in every decision you make unless you deliberately account for it.

The practical response is to measure it the only way it can be measured: directly. Ask the assistants the questions your guests ask, on a regular cadence, and record whether you appear, what they say about you, and whether the facts are correct. It is manual. It does not produce a chart automatically. And it is the only honest window you have into a channel that is increasingly deciding which hotels travelers consider. Treat that check as a real metric, put it in the report, and you will at least be looking at the thing that is happening rather than only at the things your analytics happen to be able to count.

The booking-engine handoff: hospitality's unique disaster.

Every industry has attribution problems. Hotels have an additional one that is close to unique, and it is severe enough that many properties have effectively no reliable revenue attribution at all without realizing it.

Here is the mechanism. A guest reads your content, browses your rooms, and clicks "Book Now." At that moment they are frequently handed off — to a third-party booking engine, often on a different domain. The transaction, the conversion, and the revenue all happen over there, on a system you do not own. If that handoff is not carefully instrumented, one of several bad things happens.

The booking engine may be treated as a separate site, so the session breaks. The guest arrives at the booking engine looking like a brand-new visitor from your website — a referral — and the entire history of how they found you (the blog post, the search, the AI recommendation) is severed. The booking gets credited to "yourhotel.com" as a referrer, which is a spectacularly useless piece of information. Alternatively, the revenue may be recorded only in the booking engine's own reporting and never flow back into your analytics at all, so your website reports traffic and your booking engine reports revenue and no one can connect the two. Or attribution data may simply be dropped in the handoff, so the conversion is recorded but its origin is not.

The consequence is that many hotels can tell you how much traffic their content earned and how much revenue their booking engine produced, but cannot connect the two — which means they cannot tell you what their content earned. And a marketing channel whose revenue contribution cannot be demonstrated is a marketing channel that eventually gets cut, regardless of how well it is actually working. This is how good SEO programs die: not because they failed, but because nobody could prove they succeeded.

Fixing this is unglamorous, technical, and among the highest-value things a hotel can do with its analytics. It generally means ensuring the booking engine is properly configured to preserve the session and the attribution data across the handoff — passing the tracking context through, excluding the handoff from being counted as a referral, and making sure completed bookings and their revenue flow back into your analytics tied to the original source. Your booking-engine vendor should be able to help with this, and their willingness and ability to do so is a meaningful signal about the vendor. Ask them directly how attribution and session continuity are preserved through their flow, and whether booking revenue can be passed back with the original source intact. It is the same vendor conversation that governs booking engine crawlability, and it deserves the same insistence.

It is worth pausing on how counterintuitive this is, because it inverts the instinct most marketers have been trained on. We were taught that better tracking produces better decisions — that the answer to uncertainty is more instrumentation. But when the instrument is systematically biased rather than merely noisy, more faith in it produces worse decisions, not better ones. A broken speedometer that always reads high is more dangerous than no speedometer at all, because you will trust it. That is the situation most hotels are in right now, and the first step is not a new tool. It is recognizing which of your numbers have quietly stopped meaning what you think they mean.

The OTA billboard problem.

There is a second attribution distortion specific to hotels, and it cuts in the opposite direction — it makes a channel look better than it is.

The well-known "billboard effect" describes what happens when a traveler discovers a hotel on an OTA, then leaves the platform to research the property directly — reading your website, checking your photos, looking at your reviews — and finally books. Depending on where they book and how the journey is tracked, the credit can land in places that badly misrepresent what happened. If they return to the OTA and book there, the OTA claims full credit for a booking that your own website's content and photography may have actually closed. If they book directly with you, the visit may appear as direct or as a referral from the OTA, obscuring the fact that the OTA was the discovery source.

The practical consequence is that the relationship between your OTA channel and your direct channel is far more entangled than either channel's reporting suggests, and both channels overstate their independence. Your OTA commission report shows bookings the OTA "produced," some meaningful fraction of which were actually persuaded by your own site. Your direct bookings include travelers who first found you on an OTA. Neither number is a clean measure of what that channel independently contributed.

This matters when you are making the case for direct-booking investment, because the naive comparison — OTA bookings versus direct bookings — flatters the OTA. A more honest framing looks at the total picture and asks what shifts the mix over time: as your own visibility, content, and direct path strengthen, does the share of bookings arriving directly grow? That trend, tracked over quarters, tells you more than any single attributed booking, and it connects to the economics laid out in direct booking versus OTA economics. The goal is not to win an attribution argument; it is to move the mix.

Why "direct" is the most misleading number in your report.

Let's dwell on the specific trap, because it drives more bad hotel marketing decisions than any other single metric.

"Direct" traffic is supposed to mean a visitor who typed your URL or used a bookmark — a strong brand signal. In practice, it has become the residual category: everything analytics cannot classify. That now includes a growing share of AI-referred traffic, broken sessions from the booking-engine handoff, stripped referrers from messaging apps and email clients, privacy-blocked visits, and untagged campaigns.

So when a hotel sees "direct" traffic climbing, the natural reading — our brand is getting stronger, people are coming straight to us — is often exactly backwards. What is frequently happening is that traffic earned by content, by AI citations, by search, is being stripped of its source and dumped into direct. The channel is not growing. The measurement is failing. And because the failure produces a flattering number rather than an alarming one, nobody investigates. A metric that lies in a comforting direction is far more dangerous than one that lies in a frightening one.

The strategic damage is precise and severe: organic looks flat, direct looks strong, and the rational-seeming decision is to cut the SEO and content investment that is, in reality, generating the very traffic being credited to direct. The hotel then dismantles the machine producing its results while congratulating itself on its brand. I have watched this happen. The numbers were all real. The conclusion drawn from them was catastrophically wrong.

What you can actually fix.

You cannot restore perfect attribution — that world is gone and is not coming back. But you can dramatically improve the honesty of your picture, and most hotels have substantial, unclaimed gains available. Here is the order I would work in.

A diagnostic you can run this week.

Before you fix anything, it is worth finding out how badly broken your picture actually is. Here are four checks that cost nothing and will tell you, quickly, whether you are making decisions on fiction.

Check one: what fraction of your traffic is "direct," and is it growing? Look at direct as a share of total sessions over the last two years. For most hotels, a large and growing direct share — particularly one that grows without any corresponding brand campaign, PR, or offline activity that would explain it — is a strong indication that unattributed traffic is accumulating there. If direct is your biggest channel and you cannot explain why, you have found your problem.

Check two: does your own domain appear in your referral report? If it does, your session is breaking somewhere — most likely at the booking-engine handoff, where the guest leaves your site, arrives at the engine, and gets recorded as a fresh visitor referred by you. This is one of the most common and most damaging misconfigurations in hotel analytics, and it is visible in about thirty seconds.

Check three: can you connect a single booking to the content that produced it? Pick a booking. Try to trace it back through your analytics to the source that originated the journey. If you cannot — if the trail ends at "direct" or dies at the booking engine — then you do not have revenue attribution, and every claim anyone makes about channel ROI at your property is an assertion, not a measurement.

Check four: does your search console show growth your analytics does not? Compare your impressions and clicks for non-branded, high-intent queries against what your analytics reports as organic traffic. Persistent, growing divergence between the two is a signature of attribution loss — search is delivering, and your analytics is filing the arrivals somewhere else.

Any one of these coming back badly is enough to justify the remediation work below. All four coming back badly, which is common, means your marketing decisions are currently being made on numbers that bear only a loose relationship to reality — and the most dangerous part is that the numbers look completely plausible.

The measurements that survive.

Once you accept that session-level attribution is permanently degraded, the useful move is to shift weight onto measures that are robust to it. These are less precise and far more honest, and they are what I would put in front of an owner.

Search visibility itself. Impressions, rankings, and query coverage in your search console reflect what search is doing for you regardless of whether the resulting traffic gets attributed correctly. If your visibility for high-intent queries is climbing, the channel is working, whatever your analytics claim about direct.

Branded search volume. This is a subtle and powerful one. When your content, your AI citations, and your visibility do their job, one of the effects is that more people search for your hotel by name — they discovered you somewhere, and now they are looking for you specifically. Rising branded search is a strong, hard-to-fake signal that your upper-funnel work is landing, and it is measurable in your search console even when the discovery itself was invisible. It is the closest thing to a fingerprint that dark traffic leaves behind, and it is covered further in hotel branded search.

AI citation presence. Whether AI assistants actually recommend you, checked by asking them the questions your guests ask. This is manual and crude and it is real evidence of visibility that no analytics platform will show you — the approach described in how Claude, ChatGPT, and Perplexity cite hotels.

Total direct-booking revenue, over time. The blunt aggregate. If your direct bookings and direct revenue are growing while your OTA dependence shrinks, the program is working — even if you cannot draw a clean line from any individual blog post to any individual booking. Sometimes the honest answer is that the machine is producing and the wiring is too damaged to say precisely which part did it.

Ask the guest. The most underrated attribution tool in hospitality is a single question at booking or check-in: how did you hear about us? It is imperfect, self-reported, and biased — and it routinely surfaces sources your analytics never saw at all, including the AI assistant that recommended you and the blog post someone read a month ago. In an era of broken digital attribution, the oldest method in the business has become newly valuable.

The lag problem: SEO's returns arrive late and get credited elsewhere.

Attribution's failures compound with a second problem that is particular to search and content: the returns arrive on a long delay, which makes them easy to disown even when they are visible.

Content published in January may not rank meaningfully until spring, may not be cited by an AI assistant until summer, and may generate its first meaningful bookings in a season when whoever approved the investment has moved on to other concerns. Meanwhile, a paid campaign run last week produces a number today. When a marketing team compares those two on a monthly report, the paid campaign wins every time — not because it produced more value, but because it produced it inside the reporting window.

This creates a systematic bias against exactly the investments that compound. Paid media is instantly attributable and stops producing the moment you stop paying. Content and search visibility are poorly attributable and keep producing for years. A measurement culture that rewards what it can see, on the timescale it looks, will reliably starve the second in favor of the first — and will call that decision data-driven while it does so. The channel comparison in hotel marketing channels makes the same point from the cost side; here it is the measurement side of the identical trap.

The remedy is to evaluate content and search on a timescale that matches how they actually work — quarters and years, not weeks — and to be explicit with owners that this is a deliberate choice rather than an excuse. A hotel that judges its content program on a thirty-day attributed-revenue number has, in effect, decided to fund only the marketing that pays off within thirty days. That is a legitimate strategy. It is not the strategy most independent hotels think they are pursuing, and it is not the one that builds a durable direct channel.

Modeling instead of counting.

There is a broader shift worth understanding, because it explains why sophisticated marketers are increasingly comfortable with less precision, not more.

The old model was counting: track every user, follow every path, assign every conversion to its cause. That model is dying, and no amount of tooling will resurrect it. The emerging model is modeling: accept that you cannot see every journey, and instead reason about relationships in aggregate — does organic visibility lead direct bookings by some interval? Does branded search rise after content investment? Does direct revenue grow when AI citation presence grows? Do bookings soften when content production stops?

For a hotel, the practical version of this is not a data-science project. It is a habit of mind: look at the aggregate relationships over months rather than demanding a clean line from click to booking, and treat convergent evidence — visibility up, branded search up, AI citations present, direct revenue up, OTA share down — as a stronger case than any single attributed number. A collection of imperfect signals all pointing the same way is more trustworthy than one precise-looking number that happens to be wrong.

And there is a genuinely useful experiment available to any hotel with the nerve: change one thing meaningfully and watch what happens over a long enough window. Stop publishing for a quarter and see what softens. Invest in one content cluster and watch its visibility, its branded search, and its bookings. Incrementality — did this change actually cause a difference? — is a harder question than attribution but a far more honest one, and it is the question that actually matters.

One caution about incrementality tests, since I have seen them run badly. The window has to be long enough to capture how the channel actually works. Pausing content for three weeks and concluding it does nothing is not an experiment; it is a demonstration that content takes longer than three weeks. Search and content operate on a lag of months, so an honest test runs on that timescale — which is inconvenient, and is also simply the nature of the thing being measured. Test on the channel’s clock, not on the reporting calendar’s.

A worked example: the report that lied.

Return to the hotel from the opening and watch the whole failure play out with the mechanism visible.

The property invested in a serious content program — destination guides, a real FAQ library, genuinely useful pages answering the questions guests ask. Over eight months, several things happened that nobody could see in the report. AI assistants began citing the property's content when travelers asked about the neighborhood, sending visitors who arrived with no usable referrer and were filed as direct. Travelers who found the guides on a phone during their research came back on a laptop days later to book — a broken cross-device journey, credited to direct. And the booking engine handoff, never properly instrumented, severed what little attribution survived, so the revenue landed in the booking engine's report with no source at all.

What the monthly report showed: organic sessions roughly flat. Direct traffic up sharply. Content program: no demonstrable revenue.

What was actually happening: the content program was working well. It was earning citations, driving discovery, and generating bookings — every one of which was being credited to a channel called "direct" that does no marketing and has no budget. The single most productive investment the hotel had made was invisible in the very report used to evaluate it.

The tell was there, if anyone had known to look. Branded search volume in the search console was climbing steadily — more and more people searching for the hotel by name, because they had encountered it somewhere. Impressions for high-intent destination queries were up substantially. The AI assistants, when asked, recommended the property by name and cited its guides. Every one of those signals said the content was working. None of them appeared in the analytics dashboard that drove the decision.

The hotel cut the content budget. Six months later, the direct traffic that had looked so healthy began to sag, and nobody could say why — because the thing that had been generating it had been quietly switched off, and the report had never once mentioned its name.

The report I'd actually build.

If I were designing the monthly picture for an independent hotel today, knowing what is broken, I would build it in three tiers — and I would explicitly label which tier each number belongs to, so nobody mistakes an estimate for a fact.

Tier one: things measured directly and reliably. Search visibility — impressions, clicks, and average position for your priority query sets, from your search console. Branded search volume and its trend. Total direct-booking revenue and room nights. OTA-channel volume and commission paid. Your owned-audience size and growth. These are solid ground. They are not attribution, but they are true.

Tier two: things observed but partially instrumented. Traffic by channel, with your improved groupings and the honest caveat that "direct" contains unknown traffic. Conversion rates through the booking funnel. AI-referred traffic to the extent you can identify it. Guest self-reported source from your "how did you hear about us" question. These are useful and directional, and they should be presented as such rather than as precise.

Tier three: things inferred. The relationships — does branded search rise after content investment? Does direct revenue follow visibility gains with a lag? Did the mix shift toward direct as the direct channel strengthened? These are arguments, not measurements, and they should be made as arguments, with the evidence shown.

A report structured this way is more honest than a conventional dashboard and, in my experience, more persuasive — because it does not require anyone to pretend that a number is more solid than it is. It also makes the fixable problems visible: if tier two is mostly noise because your booking-engine handoff is broken, that shows up immediately as a thing to go fix rather than a thing to quietly work around.

What to tell your owner.

If you are the one who has to defend a search and content investment to an owner or asset manager who wants a clean attributed number, you need a way to be honest without sounding evasive. Here is the framing I would use.

Be direct that attribution is degrading industry-wide and that this is not a failure of your reporting but a change in how the internet works — browsers, privacy protections, and AI-mediated discovery have collectively broken the tidy click-to-conversion chain that everyone relied on. Explain specifically that "direct" has become a dumping ground for unattributed traffic, so a rising direct number is as likely to indicate measurement failure as brand strength. That single point, understood, changes how every subsequent number is read.

Then present the convergent case rather than a single number: search visibility, branded search growth, AI citation presence, total direct revenue, and OTA share. Show them moving together. And be equally candid about the limits — you cannot draw a clean line from a specific post to a specific booking, and anyone who claims they can, with today's data, is either mistaken or selling something.

Finally, fix what you can fix and say so. The booking-engine handoff, campaign tagging, and AI channel groupings are real, concrete improvements that make the picture measurably more honest. An owner is generally willing to accept imperfect measurement from someone who clearly understands exactly why it is imperfect and is visibly working to improve it. What erodes trust is false precision — and false precision is precisely what a naive reading of the standard report delivers.

Consent, privacy, and the limits of what you should want.

A brief word on the ethics, because the temptation in an article like this is to treat privacy protections purely as an obstacle to be routed around, and that is both wrong and strategically shortsighted.

The erosion of tracking is not an accident or an attack on marketers. It is the result of people, browsers, and regulators concluding that pervasive cross-site surveillance was not a reasonable price for a personalized internet. You may find the consequences inconvenient — I do — but the direction reflects a genuine shift in what people are willing to accept, and a hotel that responds by hunting for ever more invasive workarounds is picking a fight it will lose while damaging the trust its direct relationships depend on.

The better response is the one that happens to also be the strategically correct one: build measurement on relationships people have knowingly chosen to have with you. A guest who joins your list, books directly, or tells you how they found you is giving you information with their consent, and that information is more durable, more accurate, and more defensible than anything you could have inferred by following them around the web. This is exactly the convergence described in first-party data and the owned audience: the same shift that makes your marketing more resilient also makes your measurement more honest. Consented, first-party measurement is not a consolation prize for the loss of tracking. It is the more solid foundation, and the sooner a hotel builds on it, the less it will suffer as the rest continues to erode.

The models themselves are a choice — and nobody told you that.

One more layer of distortion sits underneath everything above, and most hotel marketers have never been told it exists. The attributed numbers in your report are not observations. They are the output of an attribution model — a rule your analytics platform applies to decide who gets credit — and the rule was chosen for you, usually by default.

The common rules are easy to state. Last-click gives all credit to the final source before the booking, which systematically flatters whatever sits at the bottom of the funnel — branded search, retargeting, paid ads on your own name — and erases everything that created the demand in the first place. First-click does the opposite, crediting discovery and ignoring what closed. Linear spreads credit evenly, which is tidy and almost certainly wrong. Position-based weights the first and last touches heavily and squeezes the middle. Data-driven models attempt to infer contribution from patterns across many journeys, which is the most defensible approach and also the most dependent on exactly the tracking data that is now degrading.

Notice what this means for a hotel. Under a last-click model — the default in many setups — a guest who read three of your destination guides, was recommended by an AI assistant, and finally searched your hotel by name before booking will have that booking credited entirely to branded search. Your content gets nothing. Your AI visibility gets nothing. The channel that captured the demand takes the credit for creating it. Then someone looks at the report and concludes that branded search is the hotel's most efficient channel, which is a bit like concluding that the cashier is the most productive employee in the restaurant.

The practical guidance is not to obsess over choosing the perfect model, because none of them is right and the underlying data is broken anyway. It is to know which model you are looking at, and to understand which channels it structurally over- and under-credits. If your report runs on last-click, then every upper-funnel investment you make — content, destination guides, AI visibility, PR — will be systematically undervalued by design, before dark traffic even enters the picture. Combine a last-click model with the misattribution problems above and you have a reporting system almost perfectly engineered to make content look worthless. That is not a data problem. That is a configuration you inherited and never questioned.

Common mistakes to avoid.

• Treating rising "direct" traffic as brand strength. It is at least as likely to be measurement failure — AI referrals, broken sessions, and stripped referrers all landing in the residual bucket. Investigate before you celebrate.
• Cutting the channel that can't prove itself. Under broken attribution, the most under-credited channel is often the most productive one. Defunding based on attributed numbers alone systematically destroys the work that is hardest to measure — which, right now, is content and AI visibility.
• Ignoring the booking-engine handoff. If session and attribution data don't survive the handoff, you have no reliable revenue attribution at all, and everything else you measure is decoration on a broken foundation.
• Leaving your own campaigns untagged. A meaningful chunk of most hotels' "direct" traffic is their own marketing arriving unidentified. It is free to fix and embarrassing to leave broken.
• Demanding false precision. Insisting on a clean attributed number for every booking, in an environment where that number cannot honestly be produced, guarantees you will be given a fabricated one — and you will make decisions on it.
• Never asking the guest. "How did you hear about us?" surfaces sources no analytics platform can see. It is imperfect and it is data you currently do not have.

The asymmetry that should worry you.

Let me close with the strategic point, because it is the one that should change what you do on Monday.

Attribution does not fail evenly. It fails hardest against the channels that are hardest to see — content, organic search, AI visibility, the long slow work of becoming the property people know about. And it fails least against the channels that are easiest to see: paid media, which arrives tagged, tracked, and immediately attributable. This means the measurement environment has a built-in bias, and the bias runs consistently in one direction: toward the channels you rent and away from the channels you own.

A hotel that follows its dashboard faithfully, without understanding this, will drift year by year toward paid acquisition and away from owned visibility — not because that is the better strategy, but because it is the strategy the instruments can see. And the endpoint of that drift is a property that pays for every guest it gets, has no durable organic presence, and finds itself back in exactly the dependence on paid intermediaries it was trying to escape. The OTAs, note, have always been happy to be the channel with the clean numbers.

The properties that will do well over the next few years are the ones that recognize the bias and correct for it deliberately — that fix what can be fixed, measure honestly what can be measured, and refuse to defund a working channel simply because their instruments have gone partially blind. That takes a certain nerve, because it means telling an owner that the number on the report is wrong and explaining why. It is, nonetheless, the correct call. The alternative is to be perfectly data-driven all the way to a bad outcome.

Frequently asked questions.

Why is my direct traffic going up if I haven't done anything to drive it?

Because "direct" has become the default bucket for traffic your analytics cannot classify — and that increasingly includes visitors referred by AI assistants, sessions broken by the booking-engine handoff, cross-device journeys, links from messaging apps and email clients that strip referrer data, privacy-blocked visits, and your own untagged campaigns. Some of your direct traffic is genuinely people typing your URL. A growing share of it is traffic that was earned somewhere else and lost its fingerprints on the way in. Rising direct is a prompt to investigate, not to celebrate.

Is AI traffic really being misattributed as direct?

Frequently, yes. Referrals from AI assistants and chat interfaces often arrive with absent, inconsistent, or unrecognized referrer information, and default analytics channel groupings — written before this traffic existed — don't classify them as AI-driven. The result lands in direct, referral, or unassigned. You can improve this by building custom channel groupings for the AI sources you can identify, which converts an invisible channel into a partially visible one. You won't catch all of it, but partial visibility beats none.

What's the single biggest attribution fix for a hotel?

The booking-engine handoff, without much competition. If the session and its attribution data don't survive the transition to your booking engine, you can't connect any traffic to any revenue — which means you cannot demonstrate what your content or search work actually earns, which means it eventually gets cut. Ask your vendor directly how session continuity and source attribution are preserved through their flow, and whether booking revenue can be passed back with the original source intact. Their answer tells you a lot about the vendor.

How do I prove SEO's value if I can't attribute bookings to it?

Build a convergent case rather than chasing one number. Show search visibility and impressions for high-intent queries, growth in branded search volume (a strong signal that upper-funnel discovery is working), presence in AI assistant recommendations, total direct-booking revenue over time, and declining OTA share. When those move together, the case is strong even without a clean click-to-booking line. Pair it with incrementality thinking: what actually changed when you invested, and what softened when you stopped?

Isn't "how did you hear about us?" hopelessly unreliable?

It's biased and imperfect — people misremember and answer carelessly. It's also the only instrument you have that can capture the AI assistant that recommended you, the blog post someone read six weeks ago, and the friend who mentioned you at dinner. In an era where digital attribution has genuinely broken, self-reported attribution has become newly valuable precisely because its failure modes are different from your analytics' failure modes. Use it as one signal among several, not as gospel — but do use it.

Will attribution get better as the tools improve?

Not in the way people hope. The erosion is being driven by browser defaults, privacy regulation, user choice, and a discovery layer (AI) that fundamentally doesn't pass referrers the way links do — none of which is a bug awaiting a patch. The realistic path forward isn't restored precision; it's better instrumentation of what you can control (the handoff, tagging, channel groupings), heavier reliance on first-party data and identified relationships, and a shift from counting every path to modeling aggregate relationships and testing incrementality. Marketers who are waiting for the old clarity to return will be waiting a long time.

How do I know if my analytics is actually broken?

Four quick checks. Is direct a large and growing share of your traffic with no brand campaign or offline activity that would explain it? Does your own domain appear in your referral report (a sign your session is breaking at the booking handoff)? Can you trace a single actual booking back to the content or search that originated it? And does your search console show growth in impressions and clicks that your analytics doesn't reflect as organic traffic? Any one of those failing is a real problem. All four failing — which is common — means your decisions are running on numbers that only look reliable.

Doesn't this just sound like an excuse for SEO agencies that can't prove results?

It's a fair suspicion, and the distinction matters. An excuse sounds like "trust me, it's working, the data can't show it." What I'm describing is different: name the specific mechanisms that break attribution, fix the ones that are fixable (the handoff, tagging, channel groupings), and then present convergent evidence from instruments that do work — search console visibility, branded search growth, AI citation presence, direct revenue, OTA share. Someone who understands exactly why the measurement fails, is visibly repairing it, and shows you multiple independent signals pointing the same direction is doing the opposite of hiding. Be skeptical of anyone who cites broken attribution and then declines to fix any of it.

Which attribution model should I use?

Less important than knowing which one you are currently using and what it structurally distorts. Many hotel setups default to last-click, which credits whatever touch came immediately before the booking — usually branded search or a retargeting ad — and gives nothing to the content, guides, or AI citations that created the demand. That single default can make an effective content program look worthless. If your platform offers a data-driven model, it is generally more defensible; but whatever you use, read every attributed number with the model’s bias in mind, and never treat it as an observation when it is a rule.

My organic traffic is flat but bookings are up. What is going on?

That pattern is a classic signature of the problems in this article, and it is worth investigating rather than shrugging at. Likely candidates: your content is earning AI citations and recommendations that convert without ever registering as organic sessions; zero-click answers are delivering your information without a visit; the resulting bookings are landing in direct; and your booking-engine handoff may be severing whatever attribution survived. Check your search console for impressions and branded-search growth, ask the AI assistants what they say about you, and start asking guests how they heard about you. The traffic number is the weakest signal you have.

How much of my direct traffic is actually fake?

Not fake — misfiled. There is no universal figure, and anyone quoting you a precise percentage for your property is guessing. What you can do is estimate it for yourself: tag every campaign you control and see how much direct traffic disappears; fix the booking-engine handoff and see how much revenue suddenly acquires a source; build custom groupings for identifiable AI referrers and watch what moves out of direct. Whatever remains after those three fixes is a much closer approximation of genuine, typed-the-URL direct traffic. Most hotels are startled by how much of the bucket drains away once they look.

If you want to know what your analytics are hiding — where your traffic is actually coming from, whether your booking-engine handoff is destroying your attribution, and what your search work is genuinely earning — that diagnosis is part of every Digital Fox audit. You can see how we approach it on the services page. The most dangerous number in hotel marketing right now is a confident one, and the most valuable thing you can do is find out which of your numbers deserve that confidence.