Every autumn, marketing teams write a plan that assumes the ground under it hasn’t moved. Usually that assumption is close enough to true.
This year the ground moved in five places — and not all at once, which is precisely why it’s been easy to miss. Some of these shifts landed in December, some in spring, two arrived over the summer. Individually each looked like an operational detail to absorb. What makes them worth reading together is that they all take effect on the same season, and they all reward the same thing.
They reward brands that have decided what they are. They penalise brands still keeping their options open.
Here’s what that means in practice, and what I’d build a season around.
Narrow will outperform reach
Instagram’s “Your Algorithm” launched for Reels in December 2025, extended to Explore in April 2026, and has since rolled out to the main feed. It lets people manually set their own interests and prioritise the topics they want more of.
Alongside it, the role of hashtags changed. Adam Mosseri has been explicit that hashtags no longer boost reach — they categorise content so the system understands what it’s looking at. Labels, not levers. Discovery moved to what the platform can actually read: keywords in captions, text on screen, and the words spoken in the audio.
Put those together and the mechanism is unusually literal. Discovery stopped being something you tag your way into and became something the audience opts into — and a self-selected shortlist is brutal to generalists in a way a follower-count system never was.
Consider what someone actually types when choosing their topics. Not “business.” Not “marketing.” Something much closer to how they’d describe the problem they have this week. If your brand is about marketing, you’re not on that list. You’re the category the list filtered out on the way to something specific.
There’s a second-order effect worth knowing: the system categorises your account from your recent posts, so an account that ranges across unrelated themes reads as thematically confused and is harder to match to anyone’s declared interests. Inconsistency isn’t neutral. It’s a signal, and it isn’t a good one.
This is the oldest principle in positioning, finally issued with an enforcement mechanism. For thirty years, “narrow your positioning” was advice you could ignore at moderate cost — less memorable, still distributed. Now distribution itself is filtered by somebody’s stated preference, and breadth is what gets filtered.
What to do this season: decide what you stop covering. Not what to add — what to decline. The subject narrow enough to be somebody’s chosen topic is almost always narrower than feels comfortable, and the discomfort is the signal you’ve got it right.
Quotable will outperform ranked
On June 3, 2026, Google announced Search Generative AI performance reports in Search Console — dedicated views showing how often your URLs appear inside AI Overviews, AI Mode, and generative features in Discover. The reports cover impressions, pages, countries, devices, and dates, with data beginning May 18, 2026 and no earlier history available.
Two caveats before anyone goes looking. Google is rolling this out to a subset of websites first, so a large share of readers will open Search Console and find no Generative AI tab at all yet. And there is no click data and no query-level breakdown — you can see that you appeared, not what was asked or what happened next. Microsoft, for what it’s worth, got there first: AI Performance launched in Bing Webmaster Tools in beta on February 10, 2026.
Most people who do have access open the report once and close it, because the missing pieces are the ones they came for. That reaction is understandable and wrong. An impression inside an AI Overview means Google’s system selected your content as a source it was willing to build an answer on — and that selection runs on the same signals behind strong rankings: accuracy, topical authority, clear structure, specific relevance.
It isn’t a traffic report with a piece missing. It’s a ranked list of which of your pages are quotable, and no such list existed before this year.
It also exposes a gap most brands didn’t know they had. You can rank third and be quoted constantly. You can rank first and never be quoted once, because the page hedges everything and offers nothing liftable. A machine assembling an answer needs a sentence it can take without qualifying it. Vagueness isn’t penalised — it’s passed over, because there’s nothing to hold onto.
What to do this season: if you have the report, read your top pages by AI impressions and find the common factor. In my experience the pattern is reliably dull — they answer one question rather than covering a topic, they state a number or a price or a limit in plain text, they name who something isn’t for, and the claim sits in the first two sentences rather than the sixth paragraph. If you don’t have access yet, the work is identical; you just can’t measure it. Do it anyway, and you’ll have a baseline the day it appears.
Authorship will outperform volume
In a study published in spring 2026, Semrush collected 20,000 keywords, pulled the top ten Google results for each, and filtered to 42,000 blog pages, classifying each with the GPTZero detector. Content classified as human-written held position one 80% of the time; purely AI-generated content held it 9% — roughly an eight-to-one gap at the top. AI content appeared more often further down page one. Alongside the ranking data, a survey of 224 SEO professionals found 72% believed AI content performs at least as well as human-written content.
Two things an honest reading has to include.
First, AI detectors are unreliable. They misclassify in both directions, which puts real fuzziness into any study built on one — Semrush’s own coverage says so.
Second, the finding is contested. Ahrefs analysed 600,000 pages across 100,000 keywords with its own detector and found no relationship between AI use and lower rankings, reporting that 86.5% of top-ranking pages contained some AI-generated content and only a small fraction were purely human-written. Both studies can be true at once: AI-assisted content ranks perfectly well, and purely AI-generated content rarely takes the top spot. That reconciliation is more useful than either headline.
What survives both datasets is the distinction that actually matters — not AI versus human, but AI as production help versus AI as the source of the substance. Structure, length, variations, summaries, twenty headline options: hand them over, no penalty attaches anywhere. What loses is content that went through a machine but never through a person who knows something.
The perception gap is the part I’d watch. Those 72% aren’t wrong about what they observed; they’re wrong about the comparison, because nobody runs the counterfactual. You publish thirty AI-assisted pieces instead of four, traffic rises because you published thirty pieces, and the conclusion writes itself. The version where four human-written pieces earned more doesn’t exist to measure against. Volume masks quality for exactly as long as your competitors are also publishing volume.
What to do this season: invert the workflow. Decide the one specific thing you’re going to say, write that sentence by hand, and let the machine build structure around it. Polish was never the scarce resource. It’s free now, which makes it worthless as a differentiator.
Provenance becomes part of the brand
EU AI Act transparency obligations under Article 50 became applicable on August 2, 2026. Separately, Meta made AI disclosure mandatory across Facebook and Instagram ads with automatic detection — undisclosed AI content is an active reason for ad rejection, and third-party tools are caught through C2PA metadata whether you declare them or not.
The compliance reading is that photorealistic synthetic media needs a label. Fine, and narrower than the panic suggests: the trigger is photorealism, not AI involvement, so illustration and obvious stylisation sit outside it. On organic posts, Meta has said the label is disclosure rather than a penalty and doesn’t reduce distribution.
The strategic reading is more interesting. You are no longer choosing between labelled and unlabelled content — you’re choosing between labelling yourself and being labelled by software that has already read your file’s metadata. Provenance stops being a legal box and becomes a brand signal, in the same category as where you manufacture or who you employ. Some brands will lean into it deliberately. Others will quietly discover their entire feed carrying a label whose framing they didn’t choose.
What to do this season: decide your position on synthetic media before a platform decides it for you, and write it down as a rule the creative team works from — not a discovery someone makes when a client asks.
Front-loading will outperform reacting
Google’s core-updates documentation states that it continually makes updates to its search algorithms, including smaller core updates, and that these are not announced because they aren’t widely noticeable. That language was added to the documentation in December 2025, formalising something Google had said informally for years.
The practical consequence is that the public calendar of named updates now explains only part of the movement you see. In 2026 the named rollouts have been sparse — a Discover-specific core update in February, a broad core update from March 27 to April 8, and a May update that completed unusually fast. Third-party volatility trackers recorded elevated movement through June and July outside any announced window, which is exactly what continuous smaller updates produce. Based on the year’s cadence, several trackers expect the next broad core update in the August–September range.
That timing is the point for anyone with a season. Most seasonal planning silently depends on if something big happens, I’ll hear about it. For unnamed updates you won’t, and the one that is expected may well land inside your peak.
For businesses with a real season this matters more than for anyone with flat demand. A tutor’s year is decided in September. A florist’s across a handful of dates. If visibility slips three weeks before the peak and nobody notices for a month, the season is over before the diagnosis arrives.
What to do this season: move the structural work in front of the peak rather than into it. Fix thin pages and tighten your strongest ones while nothing is at stake. Build seasonal content four to six weeks early and let it age. Change nothing structural during the peak, because a fix made mid-peak is made without a clean baseline. And after a confirmed update, Google’s own guidance is to wait at least a full week after the rollout completes before analysing anything in Search Console — comparing data that’s still moving produces confident wrong conclusions.
One diagnostic decides where your effort goes. A ranking drop and an AI Overview squeeze look similar in a traffic graph and need opposite responses. A core-update drop shows falling position alongside falling impressions — you were ranked lower, so the work is depth on that page. An AI Overview squeeze shows position holding while clicks and click-through rate fall — the answer was given on the results page. Rewriting the page won’t help, because ranking was never the problem. Reading the second as the first is how teams spend six months rewriting healthy pages and conclude that search stopped working.
What all five have in common
Each one removed a proxy.
Announced updates proxied for knowing your position was stable. Clicks proxied for being useful. Hashtags proxied for being findable. Publishing volume proxied for authority. The absence of a label proxied for authenticity.
Every one of those was cheap, legible, and gameable — which is why they became popular, and why they broke. What replaced them is harder and duller in each case: being specific enough that a machine can quote you and a person can choose you.
Which is why the autumn plan that works isn’t a content calendar. It’s a decision made before the calendar: what this brand is for, who it’s genuinely for, and — the part almost everyone skips — what it declines to be.
That last one used to be a nice-to-have you could postpone indefinitely. This season it’s the input every system downstream is reading. A brand that hasn’t decided gets averaged into its category, and the category average has never won a comparison.
The work is unglamorous and it isn’t new. What changed is the price of avoiding it.
ABOUT THE AUTHOR
Alina Palii – brand strategist, founder of ALPA Marketing.
She works with founders and leadership teams on the decisions that come before the marketing: what the brand stands for, who it is genuinely for, what it declines to be, and how that translates into everything the market eventually sees. Strategy first — the content, the channels and the campaigns follow from it.
10+ years in marketing and a master’s degree in the field. She has built brands from zero for AI startups and national companies, and shaped the positioning of personal brands whose audiences buy on trust rather than on price.
Ukrainian by origin, living between Dubai, Paris and Ukraine, and working across the UAE and European markets — a vantage point that matters when a brand has to hold its meaning across cultures rather than be rebuilt in each new one.
She works across categories rather than inside one. Positioning logic travels between industries even when the audience doesn’t, and the pattern recognition that comes from moving between them is often what a category-blind competitor is missing.
LinkedIn – https://www.linkedin.com/in/alina-palii/
