When the Art World Decides You Exist, Google Usually Agrees

Some of my Venice Biennale articles reached the first page of Google. Some of the pieces on Gaza, Israel, Russia and institutional power sank so deep that being ‘indexed’ began to sound like a joke. I wanted to know what the machines were doing. At the end of this article, I am asking readers in different cities and languages to test it for themselves.

During the Venice Biennale 2026, many articles on my independent Journal began ranking surprisingly well on Google and Bing. Then I looked at some of the pieces in which I had taken the greatest risks, those dealing with ANGA, Pussy Riot, Palestine, Gaza, Israel, Russia, the United States and institutional power. Some were extraordinarily difficult to find, sometimes buried around what used to be page twenty of search results. I started by wondering what I had done wrong. I ended up somewhere much less comfortable.

Venice, and the Page-Twenty Problem

The Venice Biennale 2026 made the problem impossible for me to treat as theory. I was there during the preview, moving through exhibitions as they opened, speaking directly to artists and watching political actions before the mess of those days could be cleaned up into institutional memory. At the same time, articles connected with Venice began receiving far stronger visibility on the Journal than usual, with several reaching excellent positions on Google or Bing. Google had no difficulty understanding Venice. More than a century of cultural history had already made the introductions.

What interested me were the pieces that refused to behave quite so nicely: ANGA, Israel’s participation, Palestine, Gaza, Pussy Riot, FEMEN, Russia and the United States. These were often the articles where my original material was strongest because I had actually been there. ANGA and other cultural groups organised a strike on 8 May during the Biennale opening week in protest against Israel’s participation and over conditions of cultural labour. ANGA later said thirty national pavilions had partially or fully closed and more than 3,500 people marched towards the Arsenale. The Art Newspaper, reporting while events were unfolding, counted around eighteen full or partial closures. I prefer the disagreement to a falsely perfect number. Somewhere inside that discrepancy, a substantial part of the world’s most prestigious contemporary-art exhibition was being disrupted by political action.

The preview was already combustible. Nadya Tolokonnikova and Pussy Riot, alongside FEMEN, confronted Russia’s return to the Biennale and forced the Russian Pavilion into a public argument about Vladimir Putin, Ukraine and what national representation means during war. Reuters and The Art Newspaper covered the protests, and my later interview with Tolokonnikova grew directly from that encounter. Israel was facing protests of its own, while Palestinian artists and organisations occupied a far less protected institutional position. Once war entered the Giardini, the national pavilion system looked less like a charming historical eccentricity and more like what it had always partly been: a machine through which states perform themselves culturally.

Faisal Saleh, founder of Palestine Museum US, then made the story worse in the best possible way. We spoke about Gaza – No Words, the Venice exhibition translating images of Gaza into Palestinian tatreez. He attacked what he saw as failures of Western media and the obscenity of curating catastrophe while catastrophe was still unfolding, then turned part of the criticism towards ANGA itself, arguing that Palestinian voices could be excluded even inside movements publicly claiming solidarity with Palestine.

I could have reduced him to the Palestinian curator criticising Israel. It would have been cleaner, more convenient and politically easier to package. Instead he remained a Palestinian curator criticising Israel and then criticising people claiming to act for Palestine, because reality becomes propaganda the moment everybody is forced into the role prepared for them. If art writing decides who is good before the interview begins, it is not criticism. It is public relations with better vocabulary.

The Journal itself sits in an awkward position. I run it independently from Italy, mostly in English, with almost seven hundred articles on artists, exhibitions, performance, ecology, feminism, artificial intelligence, bodies, sexuality, war and politics. It is not owned by a museum, university, foundation, gallery, media group or corporation. There is no billionaire patron behind it and no institutional machine transferring decades of accumulated authority to every new URL before anyone has read a word.

I am not outside the system either. Artists, institutions and independent art platforms have linked to, cited and archived my work, and some reporting has travelled well beyond the Journal. That awkward middle position is useful here: enough recognition for differences in visibility to become measurable, nowhere near enough inherited authority to know in advance where a politically difficult article will land.

Independence means I do not ask anyone which side I am permitted to offend. I can write about Israel and Gaza, Vladimir Putin and Russia, Donald Trump and American power, and leave Faisal Saleh’s criticism of ANGA exactly where he put it. I do not owe these subjects symmetry, reassurance or political hygiene. If the facts collide, they collide, and if somebody wants to dismantle the conclusion they can do it after publication rather than having me perform the demolition for them.

I also did all the boring things independent publishers are constantly told will save us: titles, metadata, schema, internal links, images, categories, search intent, backlinks, rewrites and indexing checks. Many Biennale articles on the Journal rank extremely well. Some politically difficult ones have required travelling through page after page of Google or Bing until the entire concept of being ‘indexed’ begins to sound like a joke. At first I blamed myself, exactly as the system trains independent publishers to do. I rewrote titles, strengthened internal links, checked indexing, inspected backlinks and reopened Search Console often enough for the exercise to begin resembling a durational performance by Tehching Hsieh.

Eventually the ritual became ridiculous. If one article on the same Journal, by the same writer and around the same Biennale could perform beautifully while another politically difficult piece acquired lead boots, perhaps the interesting question was no longer whether I had chosen the wrong keyword. By then the problem was no longer SEO. Readers increasingly arrive after machines have already arranged the room. Google selects the first screen, Bing performs another calculation, recommendation systems decide what is likely to circulate, SafeSearch may already have classified the body in an artwork, and generative AI is beginning to decide which fragments of the web deserve to survive inside an answer. A writer can value presence, first-hand reporting, an unexpected answer, a contradiction that destroys the press release. Machines need signals. That is where the trouble begins.

Google Arrives Late

Microsoft is unusually candid about the signals search engines use. Bing explains quality and credibility partly through authority, reputation, links, discourse, transparency and first-hand information. The engineering is real, but the important thing happens before the engineer touches it because authority is manufactured socially first. People create it, money accelerates it, institutions preserve it and repetition eventually makes the whole construction look natural.

Jeff Koons does not need to introduce himself to Google. Museums, mega-galleries, auctions, collectors, books and decades of press have already done it for him. Damien Hirst arrives with another informational continent behind him, while Hans Ulrich Obrist arrives through institutions, biennials, interviews, publications and an almost absurd quantity of recorded cultural presence. Google meets none of them as an isolated name. It meets everything the art world has already built around the name.

François Pinault makes the mechanism almost physical. Private taste becomes Palazzo Grassi, Punta della Dogana, exhibitions, catalogues, archives, jobs and permanent cultural presence in Venice. Sheikha Al-Mayassa bint Hamad bin Khalifa Al-Thani sits where Qatar Museums, collecting, state policy and cultural diplomacy meet, with Qatar’s museum building, presence in Venice and Art Basel Qatar making culture part of the country’s geopolitical language. Bernard Arnault and the Fondation Louis Vuitton, Maja Hoffmann and LUMA Arles, Larry Gagosian, David Zwirner, Iwan Wirth, Manuela Wirth and Marc Payot occupy different parts of the same ecosystem of conversion. UBS collects art, sponsors Art Basel and helps produce one of the reports through which the market explains itself to itself. Ken Griffin and Steven Cohen turn private purchases into public mythology through prices so large that they enter databases, headlines and the biography of the artwork. Price becomes information, and information lasts.

This is how power becomes machine-readable. A private decision acquires a building, the building produces exhibitions, exhibitions produce documentation, journalists and scholars produce more of it, galleries preserve biographies, universities add citations and auction databases preserve prices. Eventually Google arrives, looks at the enormous pile of evidence and calls it authority. It does not need to decide whether Pinault matters. The world has already left millions of clues.

Yesterday’s visibility becomes part of today’s reason for granting more visibility, and today’s visibility becomes tomorrow’s evidence that the visible deserved to be visible. I sometimes think search engines suffer from an automated Stendhal syndrome. Put enough museums, biennials, billionaire collections, university citations, auction records and authoritative links in front of them and accumulated prestige begins to resemble proof of importance itself. The algorithm measures the mountain without asking who built it, who paid for the road, or why somebody else’s mountain never appeared on the map.

Search Is Not Neutral Just Because Nobody Is Touching the Buttons

Safiya Umoja Noble gave us one of the clearest reasons to stop treating search as an innocent window. Algorithms of Oppression demonstrated how commercial search could reproduce racist and sexist hierarchies, particularly around Black women and girls. Search does not float above society; it ingests society. Historical prejudice enters, commercial incentives enter, institutional authority enters, reputation enters, and what comes back appears on a clean white screen looking strangely objective.

Nobody has to sit behind the screen manually approving every injustice because scale can do the work perfectly well. A 2026 audit by Íris Damião, Paulo Almeida, João Franco, Nuno Santos, Pedro C. Magalhães and Joana Gonçalves-Sá examined four search engines and two language models around European and US elections and found measurable political skews. Parts of the European search environment disproportionately surfaced far-right entities; in the United States, Google gave greater prominence to subjects considered important by Republican voters, while other engines produced different skews. Different architectures can produce different political realities.

Contemporary art adds another layer of absurdity because art has spent more than a century destroying precisely the categories automated classification requires. My Journal contains naked bodies, queer bodies, feminist performance, sexuality, FGM, blood, mutilation, violence, war and trauma. Google says SafeSearch automatically identifies sexual material, nudity, violence and gore while attempting to distinguish artistic, scientific, documentary and educational contexts. Read that from inside contemporary art and the problem becomes almost comic. A machine has to distinguish pornography from feminist performance, sexual exposure from political vulnerability, blood celebrating violence from blood documenting violence, and mutilation as spectacle from an artwork condemning mutilation. I know curators who would need three essays, a symposium and dinner afterwards before agreeing on some of those distinctions. Google has a classifier. The artwork does not have to disappear completely for classification to matter because a probability only has to move.

When Politics Enters the Search Box

Then Israel started appearing everywhere I looked. In 2024, WIRED documented an Israeli government advertising campaign around searches connected with UNRWA. According to analytics from UNRWA USA’s advertising account, Israeli government advertisements appeared 44 percent of the time when both campaigns were eligible across more than 300 related search terms during the analysed period. The Israeli Government Advertising Agency paid for advertising capable of directing people searching for UNRWA towards Israeli government material attacking the organisation. Someone types a question into Google because they want to understand a politically contested humanitarian organisation and a government has already bought part of the answer space.

The interface looks like information, but underneath it is also an auction. Google’s relationship with the Israeli state extends much further through Project Nimbus, the roughly $1.2 billion cloud contract awarded to Google and Amazon. WIRED, working with Israeli government documents and public statements, reported that the Israel Defense Forces had been central to Nimbus from its development, helped shape requirements and became important users of the infrastructure. Israel’s Ministry of Finance described Nimbus as supplying cloud services to the government, Ministry of Defense and IDF, while Gaby Portnoy, head of the Israel National Cyber Directorate, publicly credited the Nimbus public cloud with battlefield impact.

Google says the work is not directed at ‘highly sensitive, classified, or military workloads relevant to weapons or intelligence services.’ Meanwhile the Ministry of Finance says Defense and the IDF use it, WIRED found that the IDF helped shape it, and Israel’s own cyber chief has spoken about battlefield effects. Corporate language can continue having its argument with geopolitical reality. The infrastructure is already there.

Google is not a neutral search box floating above the world. It is a search company, advertising marketplace, AI developer, cloud provider and government contractor. People use Google to investigate wars and governments while Google simultaneously sells infrastructure inside the same geopolitical system, and Amazon through AWS and Nimbus occupies another enormous section of that machinery. At some point the fantasy that information infrastructure sits outside political power becomes harder to maintain than the alternative.

Propaganda No Longer Needs You to Read It

Then came Hanover. In August 2026, The Guardian documented an Israeli-government-funded influence operation built partly around something calling itself the Hanover Institute for Public Policy, supposedly an American think tank. It had no named staff, no obvious physical presence and apparently no legal existence capable of explaining its publishing output. Between 6 and 14 August it produced 124 quasi-academic reports totalling more than 560,000 words, many with titles looking uncannily like questions somebody might ask a chatbot about Gaza, Israel, Palestinians, genocide allegations or human-rights abuses.

This is where the information war begins to mutate. Contractors were reportedly working on favourable ‘GPT framing’ and on improving the chance that material would be encountered, retrieved or cited by generative AI. Technical traces connected parts of the infrastructure to companies involved in generative-engine optimisation, while contractors discussed getting material into repositories such as Common Crawl, one of the huge reservoirs feeding the AI ecosystem. The target was no longer only the human reader. It was the machine that might later answer the human reader.

Old propaganda wanted you to read the leaflet. New propaganda can manufacture the archive from which a machine later explains reality. An independent writer can spend three days preparing one interview, checking quotations, reading documents and attaching a real name to every sentence, while a state-backed operation produces more than half a million words of apparently institutional knowledge in nine days and structures that material around systems increasingly involved in deciding what deserves retrieval. Somebody later asks a chatbot what happened in Gaza and the machine encounters scale, apparently academic language, structured reports and volume. Truth has not disappeared. It has acquired industrial competitors.

STOIC had already shown another part of the same future. In 2024, Meta dismantled a deceptive network attributed to the Tel Aviv-based political marketing firm, and Reuters reported AI-generated or likely AI-generated material, pro-Israel content and accounts pretending to represent groups including Jewish students and African Americans. Fake identities are old. Generating an entire political environment around them with machines is something else. Artificial consensus can be produced cheaply enough to begin looking like society from a distance.

UNRWA advertising, Nimbus, Hanover and STOIC do not need to be forced into one bureaucratic diagram for the accumulation to matter. The Israeli state and organisations operating around its information strategy clearly understand that visibility is territory. They understand that the informational environment can be bought, supplied, flooded, structured and prepared before the public arrives. I see no reason to make that conclusion smaller simply because its implications are unpleasant.

Global Dominance Is Their Phrase, Not Mine

The United States is even more explicit. Donald Trump’s administration published America’s AI Action Plan in July 2025 and described the United States as being in a race for ‘global dominance’ in artificial intelligence. Not leadership, not excellence, but global dominance. The plan treats AI as an information revolution as well as an industrial one and argues that the largest AI ecosystem can set global standards while producing economic and military advantages.

Anyone tempted to accuse me of exaggerating American ambitions has a problem: the White House wrote ‘global dominance’ itself. The administration then promoted export of the entire American AI stack, including chips, cloud infrastructure, models, applications and standards. Europe meanwhile performs sovereignty in conference rooms while renting much of its digital nervous system from Microsoft, Amazon and Google. Sundar Pichai, Satya Nadella and Andy Jassy do not need European leaders to kneel physically before them because the genuflection is already built into the infrastructure. Mark Zuckerberg occupies another section of the same territory through Meta.

Trump also understands culture as power with remarkably little embarrassment. His administration moved directly into the Smithsonian fight, attacking what it described as ideological distortion and targeting exhibitions, interpretations and narratives of American history, with JD Vance involved through the Smithsonian’s governance. Apparently the United States government understands perfectly well that museums manufacture reality. It is mainly museums that continue pretending they do not. A wall label frames reality before the visitor arrives. So does a collection, a ranking and an AI answer.

The Powerful Artist Has Metadata

The art world has been training for algorithms for decades. A gallery asks where you have shown, a museum wants to know who represents you, a collector asks who else bought you, and a curator looks around to see who has already accepted the reputational risk. Everybody looks sideways before looking properly at the work. Then Google arrives, counts the institutional debris and announces that the famous artist is authoritative. Of course he is. We spent thirty years manufacturing the evidence.

Artificial intelligence may turn out to be much less revolutionary than its marketing suggests. Epistemically, it can be spectacularly conservative because it retrieves what has been documented, recognises what has been named and sees importance through previous patterns of importance. The already established arrive with museums, galleries, books, auction results, collections, citations, archives and decades of press. Another artist may have spent twenty years making work every bit as serious and leave behind almost nothing a machine knows how to respect.

Some invisible art is mediocre, and so is some art hanging in museums. Visibility settles neither question. A low ranking does not tell me whether an idea is bad; it tells me where a system placed it. The poor artist may have ideas. The powerful artist has metadata, and metadata is brutally useful because exhibitions, acquisitions, biographies, auctions, awards, catalogues and citations become machine-readable confirmations that somebody counts.

The art world loves rediscovering people it failed when they were alive. Women artists, Black artists, queer artists, artists from the wrong geography, artists who never sat at the correct dinner tables. Death is especially useful because the artist can no longer ask where everybody was forty years ago, while the museum gets to congratulate itself for courageously discovering what it previously ignored. The retrospective comes first, then the catalogue and scholarship, galleries arrive, prices correct themselves, articles multiply and Google eventually finds a mountain. Apparently genius required metadata.

Ideas suffer the same indignity. An artist can spend fifteen years working around ecology, artificial intelligence, colonialism, gender, migration, war or visibility and remain somebody vaguely ‘interested in’ those subjects. Then the correct curator gives a similar idea terminology, a museum gives it a wall, a university gives it footnotes and a major gallery gives it a catalogue. Suddenly there is a discourse. The idea may barely have changed. Its passport has.

There Is No Natural First Page

Yandex gave me one of the strangest moments in this story because some politically difficult articles of mine have occasionally performed surprisingly well there. This would make a magnificent advertisement for Russian informational freedom if Yandex itself had not been exposed manipulating politically sensitive results. A 2023 source-code leak showed mechanisms intended to prevent Vladimir Putin and Russia’s Z war symbol from appearing in embarrassing contexts. So a search engine capable of protecting Russian political power can still give one of my politically awkward Western art articles more visibility than Google or Bing gives it elsewhere. The contradiction is not an inconvenience here. It is the lesson.

There is no natural ranking and no universal first page waiting somewhere for the correct algorithm to discover it. There are architectures, and architectures classify, privilege, bury, connect and separate. Western platforms prefer cleaner vocabulary: safety, trust, quality, credibility, authority, relevance. The words sound hygienic because engineers need labels for technical systems. Their consequences do not become neutral with them.

Generative AI makes the transfer more radical because now the source itself can disappear. Pew Research Center found that Google users seeing an AI summary were much less likely to click traditional search results than users who did not encounter one, and links inside the summaries received even fewer clicks. The information arrives while the source retreats into the background. The writer becomes raw material for somebody else’s answer. For an independent publication this is not merely a traffic problem. It is a transfer of cultural power. The twentieth-century struggle concerned who owned the printing press. The twenty-first-century struggle may concern who owns the answer and, more dangerously, who managed to get inside it before the question was asked.

I Thought This Was an SEO Problem

I thought I had an SEO problem. I was wrong. What I had in front of me was an old hierarchy learning new machinery, a world in which power produces consequences and those consequences become museums, archives, prices, catalogues, backlinks, government contracts, reports, advertising campaigns, databases and oceans of text. Machines encounter those traces, accumulated traces become authority, authority comes back as relevance, relevance produces more visibility, and visibility produces more evidence that the visible deserved to be visible.

The loop closes beautifully without requiring a secret committee. Nobody needs to sit in a basement deciding which artist deserves cultural existence because the old hierarchy has already produced the archive from which the machine learns. The machine inherits the archive. The archive was never neutral. Power produces evidence of its own importance and the information system returns that evidence wearing the calm face of relevance.

Nobody has to censor me, which is almost quaint now. Leave the article online, index it, give me the little green light in Search Console and then put it somewhere almost nobody will reach. The old censor burned the page. The new system can leave the page perfectly intact and remove the audience. That is the difference between freedom to publish and power to exist in public.

Readers still interfere. They search, screenshot, send, cite, link, argue and occasionally drag something back out of obscurity, and my own Journal would not exist without those actions. But their agency begins inside a room somebody else has already arranged. Then somebody important repeats something you have been saying for years and suddenly the acoustics change: a museum contextualises it, a curator names it, a university historicises it, a gallery monetises it, a collector validates it, the market prices it, Google ranks it and AI explains it. Everybody finally hears the idea. The idea did not necessarily improve. Its authority did.

Perhaps we did not build artificial intelligence in the image of human intelligence. Perhaps we built it in the image of human hierarchy. After PageRank, Nimbus, billionaire collections, SafeSearch, AI propaganda, museums, Donald Trump, Israel, Russia and American technological dominance, I keep returning to Alberto Sordi in Mario Monicelli’s Il Marchese del Grillo: ‘Io so’ io e voi non siete un cazzo.’ Giuseppe Gioachino Belli had already given the idea a much older Roman ancestry, but Sordi made the hierarchy indecently easy to understand. I am who I am, and you count for nothing.

The extraordinary thing about power is that eventually the sentence no longer needs to be spoken. Architecture says it, museums say it, markets say it, search results say it, and AI answers may begin saying it with perfect grammar and no trace of embarrassment. The Marchese del Grillo needed aristocratic birth. The machine needs metadata.

Now Put This Article Into the Machine

I am not asking you to agree with me. I want to know where the machines put this article. The Journal occupies an unstable middle territory that makes it useful for the experiment: almost seven hundred articles, years of accumulated material, external citations and enough recognition that a new page has a realistic chance of appearing, but no museum, university, multinational publisher or billionaire transferring authority to it before publication.

Start with the exact title, then remove the easy clues and search the actual subjects: Venice Biennale algorithmic visibility, art world power Google, ANGA search ranking, Palestine Project Nimbus, AI cultural authority. Do not add my name or the Journal’s name simply to make the article easier to retrieve. I want to know what happens when the subject itself has to find the page.

Send me the search engine, exact query, approximate position, city and country, with a screenshot if you can. Device, query language, interface language and whether you were logged in or using private browsing are useful too. An English-language article published in Italy may occupy very different informational positions when searched in English, Italian, Arabic, Hebrew, Persian, Russian or another language. An English query from Tehran is not the same informational event as a Persian query from Tehran, just as an Arabic search from Ramallah is not simply an English result written in another alphabet.

I want results from New York, London, Paris, Berlin, Rome, Venice, Tel Aviv, Ramallah, Beirut, Tehran, Dubai, Doha, São Paulo, Buenos Aires, Lagos, Johannesburg, Delhi, Beijing, Seoul, Tokyo, Sydney and places I would never think to name. A first-place ranking matters, position ninety-three matters and not finding the article in the first hundred results matters. If absolutely nothing strange happens, that belongs in the result too.

I will not build a little museum of screenshots selected because they make Google look guilty. If the article performs beautifully in Tehran and disappears in London, I want both. If Bing loves it in New York and Google loves it in Berlin, I want both. If Yandex constructs another hierarchy entirely, I want that too. If my suspicion collapses, I want the wreckage.

If enough observations arrive, I will organise them by search engine, exact query, language, location, device and approximate position and publish the pattern before deciding what the pattern means. The experiment may support this article, damage it or blow a hole straight through it. Good. Google, Bing and other systems have spent years crawling us, classifying us, measuring us and deciding how likely another human being is to encounter what we produce. This time the article goes back into the machine. Tell me where it comes out. I will publish what we find.