Optimizely says your website's primary audience is no longer human, and the rebrand was the small part of the story
Optimizely unveiled a new brand identity on 22 June 2026, and the new logo got all the attention, which is the least interesting thing that happened. Twelve days earlier the company had launched a platform for measuring how AI systems read your website. Ten weeks later, at its own conference, an Optimizely executive told the room that 60% of web visits are now AI referral traffic and crawlers, and that your primary audience is no longer human. If that figure is anywhere near right, it is the most consequential claim made about websites this year, and almost nobody in the UAE market is planning for it.
This piece sets out what Optimizely actually announced, which parts are verifiable and which are vendor claims, and what a business in Dubai should do about it. The short version is that you do not need to buy anything to act on most of this.
What Optimizely actually announced, in order
The rebrand reads differently once you see it as the middle of three announcements rather than a standalone event. The sequence matters.
| Date | Announcement | Why it matters |
|---|---|---|
| 10 June 2026 | Full AEO platform launched, with a partnership with Conductor. Includes Agent Visibility Analytics built inside Optimizely Analytics | The first mainstream enterprise platform to treat AI crawler and agent activity as a measurable marketing channel with its own analytics |
| 22 June 2026 | New brand identity, built around the idea that marketers should be free to grow | The visible signal of a repositioning from digital experience platform to, in the company's own words, the AI platform for marketing |
| 1 September 2026 | Camp Opticon, with three trails and twelve announcements, including the claim about AI traffic and the renaming of the Opal assistant to Mark | Where the strategy was spelled out, including the agent audience argument and an open source benchmark |
One detail is worth noticing, because it tells you how fast this moved. The 10 June release still describes Optimizely as "the leading digital experience platform (DXP) provider". The 22 June release describes it as "the AI platform for marketing". The company changed how it describes itself in twelve days.
A note on a phrase doing the rounds. Secondhand coverage has described the rebrand as a shift to an AI-powered marketing operating system. That is not Optimizely's language and we could not trace it to the company. Its own words, in the release and on its product pages today, are "the AI platform for marketing". Small thing, but if you are citing this internally, cite what they said.
The rebrand itself, and the irony the company leaned into
The rebrand was developed with the London brand agency Ragged Edge, working with the artist Andrey Kasay, whose illustration work has previously appeared for brands including MTV, Nike, Gucci and Red Bull. The result is deliberately strange for enterprise software: parrots, flaming coffee cups and, in one widely shared asset, a man driving a hamster.
Optimizely was explicit that while AI supported elements of production, creative direction remained human led throughout. An AI company launched a brand made by human hands, and said so in the release. That is the argument of the rebrand, not a footnote to it.
Alex Atzberger, Optimizely's CEO, put it this way in the announcement: "AI should make the mundane elements of marketing easier so brands can focus on the things that AI could never replace: human curiosity, creativity, and compassion which allows them to build something distinctive."
The reaction split, loudly, which appears to have been the point. The brand was described in the release as built to be felt before it is understood, and deliberately hard to ignore in a category trending towards sameness. A partner agency reviewing it in July noted that the only reaction that kills a brand is indifference, and that this one did not get much of that.
Underneath the illustrations sits the research the rebrand was built on, and it is the part a Dubai marketing team should read. Optimizely commissioned Savanta to survey 2,003 B2B marketing leaders across multiple regions in May 2026, with roles from manager to C-level. Three findings:
- 64% worry that mass AI adoption is creating a sea of sameness. Nearly two thirds of marketing leaders think the tools everyone adopted to differentiate are making everyone look identical.
- Only 31% say AI makes them feel genuinely free to do their best work. Which means roughly seven in ten do not.
- 25% admit they regularly publish content they know is not on brand, because volume pressure outweighs the time to get it right.
That last number is the uncomfortable one. A quarter of marketing leaders are knowingly shipping work that is off brand, not through incompetence but because the volume expectation leaves no room to fix it. Full findings were still to be published at the time of writing, and the survey was commissioned by a vendor whose product addresses the problem it describes, which is worth holding in mind. But the pattern will be recognisable to anyone running marketing in this region.
A practical starting point
Find out which AI systems are already reading your website
The data is in your own server logs and most businesses have never looked. We will tell you which AI crawlers and agents are hitting your site, which pages they take most, and which of your pages they ignore entirely.
No platform purchase involved. This is a log analysis and a report.
The claim that should change your 2027 plan
The claim that matters came from the agent experience session at Camp Opticon, where Optimizely's Naz Ramezani stated that 60% of web visits are now AI referral traffic and crawlers, and that your primary audience is no longer human. Optimizely's president, Shafqat Islam, reported it as the point that should reorganise your 2027 plan.
Treat the number carefully. Optimizely published no methodology alongside it, no sample definition and no detail on which sites it is drawn from. It is a vendor figure from a company selling agent visibility tooling, and the honest position is that it is a claim rather than an established fact. The direction of travel, though, is not in dispute by anyone who has looked at their own logs recently, and that is the part you can check for yourself in an afternoon.
If even a fraction of that is true for your site, several things follow that have nothing to do with buying software.
- Your analytics has been measuring the minority. Most analytics platforms are built around human sessions and filter bots out by design. If agents are a large share of retrieval activity, your reporting has been systematically blind to it.
- Content that reads well to a person can still be unusable to a retrieval system. A claim buried in an image, a specification locked in a PDF, a key fact that only exists in a video, a price or a credential rendered by a script. A human works around these. A retrieval agent does not.
- Being cited matters separately from ranking. A page can rank respectably and still never be the source an answer engine quotes, and the reverse is also now true.
- Your competitors' content is being read on your customers' behalf. If a buyer asks an assistant to compare three Dubai suppliers, the comparison is assembled from whatever each site made machine readable. The best supplier does not necessarily win. The clearest one does.
The Conductor partnership, and the one genuinely new idea in it
The AEO platform launched on 10 June is a partnership rather than a build, with Conductor supplying AI and search performance intelligence while Optimizely supplies what it calls observed agent traffic data. Conductor's CEO and co-founder Seth Besmertnik described it as exclusively powering Optimizely's AEO data pipeline.
Most of that is commercially interesting and operationally irrelevant to a business in Dubai. One idea inside it is neither.
Optimizely's Agent Visibility Analytics works from, in the company's words, factual log level data from an organisation's own site rather than inferred or modelled behaviour. It classifies AI requests by intent, separating retrieval from indexing from training activity, and lets teams analyse which AI systems are accessing content and what information they prefer.
Here is why that matters to you. The data source is your own server logs. You already have it. Nobody needs to sell it to you, and nothing about reading it requires an enterprise platform. Shafqat Islam made the distinction himself when he argued that prompt monitoring tells you what a model says in a controlled environment, while what marketing leaders actually need to know is which agents are showing up on their sites and what content they are pulling.
That is the single most useful takeaway from the whole sequence of announcements, and it is free. The gap between a business that knows which AI systems read its site and one that does not is currently a log file and somebody willing to look at it.
Opticon, the agent platform numbers, and how much to believe
Camp Opticon produced twelve announcements across three trails, and the figures Optimizely put behind them are worth recording alongside the appropriate scepticism.
| What Optimizely reported | How to read it |
|---|---|
| More than 500,000 agent executions since the Agent Platform launched a year earlier, on track to consume a trillion tokens | Usage volume, which the company itself called a vanity metric on its own. It does establish that this is in production rather than in pilots |
| One customer marketing workflow cut by 97%, campaigns live in minutes where the old cycle was weeks, first pass brand and legal review by agents pushing output velocity up 30% | Customer outcomes selected by the vendor, unnamed and unaudited. The brand and legal review one is the most believable, because review queues genuinely are where weeks disappear |
| An in house AI lab post training open weight models, reporting 10x better cost efficiency than frontier models on the same work with quality up about 7% | A real and increasingly common engineering argument. The right sized model for a narrow task beating a frontier model on cost is well established. The specific multiple is Optimizely's own measurement |
| An open source AI Benchmark for Marketing: 285 tasks across 15 marketing functions, 6,523 expert written rubric criteria, each run scored by two independent LLM judges | Genuinely useful if the industry adopts it, and publishing it openly is a real commitment. But Optimizely built the benchmark and then reported its own platform scoring 58% against Claude Code's 52% at 2.1x lower cost. Interpret a vendor's win on a vendor's benchmark accordingly |
The three trails are the clearest statement of strategy. Marketer experience brought Virtual Teammates, AI colleagues with their own email, identity, access and personality, five of them described as hireable: Chief of Staff, SEO and AI Search Analyst, Marketing Analyst, Personalization Strategist and CRO Manager. Customer experience brought Limitless 1:1 Personalization, Autonomous Optimization and Dynamic Experiences. Agent experience brought Agent Visibility with Conductor and what Optimizely calls Agent-Ready Experiences, intended to make a site something agents can act on rather than merely read.
Islam also set a public test for the company: that Optimizely's own marketing team will double pipeline in 2027 without adding headcount, and that if the platform cannot do that internally you should not believe it will do it for you. That is an unusually falsifiable commitment from a software vendor, and it is worth revisiting in a year.
The naming lesson, which is the most copyable part of the rebrand
The naming change is the part of this rebrand that a UAE business can actually learn from, and it got almost no coverage because illustrations are easier to write about than taxonomy.
Optimizely used to wrap its portfolio in invented brand names. Optimizely One as the umbrella, Opal as the AI assistant. The products page today reads as plain description instead: Content Management System, Experimentation, Agent Platform, Content Marketing Platform, Digital Asset Management, Personalization, Analytics, Feature management, Commerce. The flagship products are now described as agentic CMS and agentic experimentation.
The logic is specific and it applies to anyone writing a website in 2026. Buyers increasingly research through answer engines, and so do the assistants doing their shortlisting. "Agentic CMS" is a phrase both a person and a model can parse instantly. "Opal" is not. So the personality moved up into the brand layer, where it can be as strange as it likes, and the product catalogue became ruthlessly literal.
The inversion is worth stating plainly because so many businesses do it backwards: a sober, forgettable corporate brand on top, and underneath it a mythology of invented product names that nobody searches for and no model can interpret.
There is a caveat that keeps the lesson honest. At Opticon, Opal did not simply disappear into plain language. It became Mark, split into Mark IQ as the data and context layer and Mark-1 as a family of purpose built models. So Optimizely has kept one bespoke name, for the layer it considers proprietary, while making everything a buyer shops for literal. That is arguably the right split rather than a contradiction, but it does mean the rule is not "never invent a name". It is closer to this.
- Name what the buyer searches for literally. Your service pages, your product categories, your capability names. If someone would type it into a search box or ask an assistant about it, use the words they would use.
- Save invented names for what is genuinely proprietary. A model, a methodology, a framework that is actually yours and that you intend to build recognition for over years.
- Never make a buyer learn your vocabulary to understand your offer. Every invented name is a translation step, and in an answer engine that translation step is where you get dropped from the comparison.
- Check what the old names are still doing. Optimizely's own products page carried a page title referring to Optimizely One months after the shift, which is a useful reminder that renaming a brand and renaming a website are different projects.
For most businesses in Dubai this is a one day content exercise with a disproportionate payoff, and it costs nothing but the decision. Our SEO services cover the search side of it, and website amendments cover putting it live.
What a UAE business should actually do about this
Almost nothing in the Optimizely announcements requires you to buy Optimizely, and we would rather say that than pretend otherwise. The useful response for a business in Dubai is eight things, and six of them cost only attention.
- Read your own server logs. Find out which AI crawlers and agents are requesting your pages, how often, and which pages they take. This is the single highest value thing on this list, it uses data you already own, and most businesses here have never once looked.
- Separate crawler traffic from human traffic in your reporting. If your analytics filters bots by default, you are not measuring the audience Optimizely is arguing has become the majority. Report both, separately, and watch the ratio over time.
- Get your facts out of images, PDFs and scripts. Prices, specifications, credentials, licence numbers, opening hours, service areas, contact details. If a fact matters and it only exists inside an image, a PDF or something rendered by JavaScript, treat it as invisible and move it into text.
- Answer questions in a form that can be lifted. A clear question as a heading and a complete, self contained answer underneath. Not because it games anything, but because a retrieval system can only quote a passage that stands on its own.
- Publish the specifics nobody else will. Real numbers, real constraints, real limitations, named people with real credentials. The sea of sameness that 64% of marketing leaders are worried about is made of content that could have come from any competitor. Specificity is the only defence, and it is also what gets cited.
- Keep your own facts consistent everywhere. Website, Google Business Profile, directory listings, social profiles, PDFs. Contradictory facts across sources make you a poor candidate for citation, because a model has no way to decide which version is true.
- Decide deliberately what you allow and block. Your robots file, and whatever position you take on training versus retrieval crawlers, is now a commercial decision rather than a technical one. Make it on purpose, with someone who can explain the trade off.
- Then, if the scale justifies it, look at tooling. Agent visibility tooling makes sense for an enterprise with thousands of pages and a team to act on it. For a Dubai SME with forty pages, the first six items on this list will deliver more than any platform, and in a fraction of the time.
The uncomfortable part. Everything above is content and technical hygiene. There is no product that fixes a website whose facts are locked in a brochure PDF and whose service pages say the same thing as every competitor's. That work is unglamorous and it is the work.
The practical version
Get your site ready for the readers that are not people
We will analyse your logs for AI crawler and agent activity, identify every fact currently trapped in an image, a PDF or a script, restructure your key answers so they can be cited, and give you a position on what to allow and block, with the reasoning.
What we think is actually going on
Our reading is that Optimizely has made a bet that is correct in direction and unproven in scale, and both halves of that sentence matter.
The correct part is the three audience model. A website has always had two readers, the person using it and the search engine indexing it. There is now a third, a system acting on a person's behalf, and it reads differently from both. Treating that as a distinct audience with its own requirements is the right conceptual move, and Optimizely is early to make it a product pillar rather than a blog post.
The unproven part is the 60%. Without a methodology it is a marketing number, and the gap between crawler requests and commercially meaningful discovery is wide. A crawler hitting your site a thousand times is not a thousand customers, and a share of web requests is not a share of revenue. Anyone reorganising a budget around that figure alone is making a mistake, and the people selling you the figure are also selling the remedy.
What is not in doubt is the underlying shift. The number of buying decisions that pass through an AI intermediary before a human ever reaches your website is growing, in this market as everywhere, and the businesses whose content is specific, structured and consistent will be represented in those answers while the businesses with beautiful, vague websites will not. That has been the argument for good content for twenty years. The difference is that the penalty for ignoring it now arrives faster and is harder to see, because you do not get a bounce rate for an answer you were left out of.
And the honest closing thought on the rebrand itself: an AI company spending real money to make something a machine could not have made, and saying so explicitly, is the most interesting strategic statement in the whole announcement. Whether you like the hamster is beside the point.
Optimizely rebrand and AI agent traffic FAQs
Common questions about the Optimizely rebrand, the agent audience claim and what it means for a business in the UAE.
What did Optimizely actually announce in June 2026?
Two things, twelve days apart. On 10 June 2026 it launched a full AEO platform in partnership with Conductor, including Agent Visibility Analytics that measures AI crawler and agent activity using log level data from a company's own site. On 22 June 2026 it unveiled a new brand identity built around the idea that marketers should be free to grow, developed with the agency Ragged Edge and the artist Andrey Kasay. The rebrand signalled a repositioning from digital experience platform to, in Optimizely's own words, the AI platform for marketing. A third wave of announcements followed at Camp Opticon on 1 September 2026.
Is Optimizely now calling itself a marketing operating system?
Not in anything we could find from the company. That phrase appears in secondhand coverage but it is not traceable to Optimizely's own announcements or product pages. Optimizely's self-description, in the rebrand release and on its products page, is "the AI platform for marketing". Worth getting right if you are citing this in a strategy document, because the two phrases imply different things. An operating system implies everything runs on top of it. A platform for marketing is a narrower and more defensible claim.
Is it true that 60% of web visits are now AI agents?
That is Optimizely's claim, made at its own conference by Naz Ramezani and reported by the company's president, and it should be treated as a claim rather than a fact. Optimizely published no methodology, no sample definition and no detail on which sites the figure is drawn from, and it is a vendor number from a business selling agent visibility tooling. The direction is real and most people who check their own logs are surprised by what they find, but the specific percentage is unverified and we would not plan a budget around it. Check your own logs instead, which takes an afternoon and tells you about your site rather than an average.
How do I find out whether AI systems are reading my website?
Your web server access logs already record it, and this is the point worth taking from the whole Optimizely announcement. Every request to your site is logged with a user agent string, and AI crawlers and retrieval agents identify themselves in that string. Analysing those logs tells you which systems are requesting your pages, how often, and which pages they take most and ignore entirely. Optimizely built a product around exactly this data source, describing it as factual log level data rather than inferred behaviour, but the data is yours and reading it needs analysis rather than a platform purchase.
Do we need to buy Optimizely or a similar platform?
Almost certainly not, and we would rather say so. Agent visibility tooling makes sense for an enterprise with thousands of pages, a dedicated team and the capacity to act on what it finds. For a UAE business with a few dozen pages, the work that will actually change your position is content and technical hygiene: getting facts out of images and PDFs, structuring answers so they can be quoted, publishing specifics instead of generalities, keeping your facts consistent across every platform, and deciding deliberately what you allow and block. None of that requires a licence, and no platform substitutes for it.
What should we change on our website first?
Get your facts into text. Prices, specifications, credentials, licence numbers, opening hours, service areas and contact details that exist only inside an image, a PDF or something rendered by a script should be treated as invisible, because to a retrieval system they effectively are. A human squints at an image and works it out. An agent does not. After that, restructure your most important answers as a clear question heading with a complete, self contained answer underneath, because a retrieval system can only quote a passage that stands on its own.
What was the research behind the rebrand?
Optimizely commissioned Savanta to survey 2,003 B2B marketing leaders across multiple regions in May 2026, with respondents from manager to C-level across a range of company sizes and industries. Three findings were published with the rebrand: 64% worry that mass AI adoption is creating a sea of sameness, only 31% say AI makes them feel genuinely free to do their best work, and 25% admit they regularly publish content they know is not on brand because volume pressure outweighs the time to get it right. Full findings were still to be published at the time of writing, and the survey was commissioned by a vendor whose product addresses the problem it identifies.
What is the naming lesson from the rebrand?
Optimizely dropped most of its invented product names, including Optimizely One as an umbrella and Opal as the AI assistant brand, in favour of plain description: Content Management System, Experimentation, Agent Platform, Personalization, Analytics. The reasoning is that buyers and the assistants shortlisting on their behalf both research through answer engines, and a literal phrase can be parsed instantly while an invented one cannot. The lesson is to name what a buyer searches for literally, reserve invented names for something genuinely proprietary, and never make a buyer learn your vocabulary to understand your offer. Optimizely kept one invented name, Mark, for its model and context layer, which is a reasonable split rather than a contradiction.
Should we block AI crawlers from our website?
It is a commercial decision rather than a technical one, and it deserves a deliberate answer rather than a default. Blocking retrieval crawlers removes you from the answers your buyers are being given, which for most businesses competing for consideration is the wrong trade. Training crawlers are a separate question and reasonable people reach different conclusions, particularly where original research, proprietary data or licensed material is involved. What we would advise against is leaving it to whatever your platform shipped with, or blocking everything out of caution without understanding which category you are blocking.
Does any of this replace SEO?
No, and the framing is misleading. Most of what makes a page usable to a retrieval agent is what has always made a page good: a clear structure, facts in text, honest specifics, consistent information and a site that loads properly. What changes is that a second kind of reader now has different tolerances, and the cost of ignoring it is invisible rather than obvious. You do not get a bounce rate for an answer you were left out of, which is precisely why it goes unnoticed. Treat it as an extension of search work rather than a replacement for it.
Related services
The services behind the work on this page.






