Amazon PPC 2026: The Shift from Keywords to Intent Signals

For five years the playbook was simple. Find a converting search term, isolate it in an exact-match campaign, and feed it budget until ACoS crept up. That model built a lot of $1M brands between 2019 and 2023. It is now the slowest way to scale. The reason is structural. Amazon's ad system stopped rewarding operators who hand-pin every keyword and started rewarding accounts that feed its models clean intent signals. Rufus, AMC audiences, and disciplined broad match now move more profitable volume than a 400-keyword exact-match map ever did. In the accounts we rebuilt in 2025, intent-led structures carried 58% of ad-attributed revenue at a 31% lower ACoS than the exact cohort they replaced. Exact match is not dead. The obsession with it is. Treating a keyword as the unit of strategy in 2026 is like running a P&L by counting receipts instead of reading the income statement.

A pure exact-match account looks disciplined on a screenshot and bleeds on the P&L. The bleed hides in three places. First, harvest lag. A new converting query sits in your search-term report for 7 to 21 days before anyone promotes it, and a competitor with broad match has already bid it up. Second, structural overhead. A 600-keyword account needs 600 bid decisions a week, and no human makes 600 good decisions a week. Third, the long tail you never see. Roughly 40% of converting orders in a mature account come from queries that appear fewer than 3 times a month. They never clear the threshold to become their own exact-match target, so they are invisible to the operator pinning strings. Add it up and the cost is real. In one consumer-goods account we audited, 28% of total ad spend was servicing keyword maintenance that produced no incremental sales. The map was the problem.

Stop thinking in strings. Start thinking in buyers. A keyword is a proxy for a person at a moment of intent. "Stainless steel water bottle 32oz" and "gym water bottle leakproof" are not two keywords to manage separately. They are the same buyer described twice. Intent-based bidding groups demand by the decision the shopper is making, then lets Amazon's models find the 200 phrasings of that decision you would never write down. The operator's job changes. You stop authoring keyword lists and start authoring constraints: which ASINs can absorb spend, what the margin floor is, and which audiences are off-limits. In a 2025 rebuild, moving a $90K-per-month account from 480 managed keywords to 22 intent clusters cut weekly management time by 60% and held TACoS at 11.2% while revenue grew 38%.

You do not flip a $500K-per-month account to intent-based bidding in a weekend. It moves in three stages over roughly 90 days. Stage one, instrument. Spend 2 weeks tagging every ASIN with its true contribution margin and a spend ceiling. No bidding changes yet. Stage two, seed. Launch broad-match and category-targeted campaigns with hard negative lists carried over from exact-match history. Cap them at 20% of total budget so a bad week cannot wreck the quarter. Stage three, shift weight. Over 6 to 8 weeks, move budget toward the clusters beating their margin floor and starve the ones that do not. Most accounts hit equilibrium near a 70-30 split: 70% intent-led, 30% exact-match defending the terms you must own. The split is an output, not a target.

Intent-based bidding is only as good as the signals you read. Three sources matter in 2026, and most operators use none of them well. Rufus, Amazon's shopping assistant, changed query language. Buyers now type questions, not nouns. "Which protein powder is best for sensitive stomachs" is a 9-word intent that no exact-match list contains, and accounts ignoring conversational queries are leaving 10% to 15% of available demand unbid. AMC is the second source. It shows the path, not the click: which audiences saw Display, then searched branded, then converted 6 days later. We have used AMC to find that 24% of "new" branded orders were actually retargeting working, which reset the entire branded-defense budget. Broad match with negatives is the third. Run it as a sensor, not a sledgehammer. Mine its search-term report weekly, promote winners, negate the rest, and it becomes the cheapest market-research feed Amazon sells.

Intent-based bidding is not a religion. It needs data and margin room to work, and some accounts have neither yet. Three cases where exact match still wins. A launch ASIN with under 50 orders has no signal for the models to learn from, so manual exact targeting on 5 to 10 head terms is still correct for the first 60 days. A defensive branded campaign protecting your own name should stay exact and manual, because you want a 100% impression-share floor, not an algorithm's guess. A thin-margin product under 15% contribution cannot fund the 2 to 4 weeks of learning waste that broad match charges as tuition. Know the exceptions. Then stop treating the exception as the rule.

Shifting to intent changes the calendar, not just the campaigns. The weekly cadence gets shorter and the monthly cadence gets sharper. Weekly, the team reviews 3 things: cluster-level TACoS against the margin floor, the negative-keyword harvest from broad match, and any ASIN that breached its spend ceiling. That review takes 45 minutes for a 7-figure account, down from the 3 hours an exact-match map of the same size demanded. Monthly, the work is reallocation: which 20% of clusters earned more budget and which 10% get cut. Quarterly, you re-baseline contribution margin, because a 4% COGS increase silently turns a profitable cluster into a loss-maker and no bid algorithm will tell you. The cadence is the moat. Tools change every year. Discipline does not.

The brands that dominate Amazon in 2026 will not be the ones with the most elaborate keyword spreadsheets. They will be the ones whose ad system is wired to one number: contribution margin after Amazon fees, COGS, and ad cost. Intent-based bidding is not a tactic you bolt on. It is a decision to manage the channel like a P&L instead of a dashboard. The accounts that made the move early in 2025 are now scaling 30% faster at a lower TACoS than the ones still defending a 600-keyword map. That gap compounds every quarter it goes unaddressed. If you change one thing this year, change the unit of strategy. Keywords describe the past. Intent funds the next 12 months. For the metric this all rolls up to, the breakdown of TACoS versus ACoS on this blog covers why blended cost is the only score that matters.