I started looking for my family in the places you are supposed to look. Ancestry. FamilySearch. The census records. I typed in the name Arceneaux and got a wall.

Not a paywall. A real wall. The 1870 kind.

The 1870 federal census was the first time the United States recorded formerly enslaved people by name. Before that, if your ancestors were Black in this country, they were listed as property. Age, sex, and color on a slaveholder’s schedule. No name. No family. No record that they existed as people.

But here is the thing about Louisiana Creoles of Color. We had it harder than that.

The wall behind the wall

Every Black American hits the 1870 wall. Louisiana Creoles hit the 1870 wall and then fall into a hole behind it. Our records are in French. Then they are in Spanish. Then they switch back to French. Then English, maybe, after 1803.

Between 1684 and 1803, Louisiana changed hands three times. French colonial rule. Spanish colonial rule. Brief French return. Then the American purchase. Each government kept records in its own language, its own legal system, its own naming conventions. A single ancestor can appear as three different people across three different archives because a French clerk wrote the name one way and a Spanish clerk wrote it another.

The Catholic Church recorded baptisms, marriages, and burials for Creole families. Those sacramental registers are some of the most important genealogical records in Louisiana. The Archdiocese will not let you see them. All requests go through the mail. No walk-ins. No digital access.

And the originals of Louisiana’s colonial census records? They are in Seville, Spain. Paris, France. Havana, Cuba. No other American state requires you to leave the country to trace your own family.

What made Louisiana different

Louisiana Creoles of Color were not just enslaved people who gained freedom. They were a distinct legal class. Under French and Spanish colonial law, the gens de couleur libres occupied a middle position in a three-tiered social order: white, free people of color, and enslaved. They owned property. They ran businesses. They had an 80% literacy rate in the antebellum South, in an era when every other Southern state was making it illegal to teach Black people to read.

These families generated records that no other Black community in America produced at that scale. Property deeds. Succession documents. Notarial acts. Manumission papers. Baptismal records. The paper trail exists. It is just scattered across four countries, three languages, and hundreds of parish churches, courthouses, and private notarial archives that have never been digitized.

The 90% problem

The commonly cited figure is that only about 10% of the world’s genealogical records are online. For Louisiana Creole records, the number is worse. The records that matter most, the church registers and notarial acts and colonial censuses, are disproportionately physical-only. They sit in drawers in Natchitoches. Behind mail-only request systems in the Archdiocese. In Spanish government databases that most American researchers do not know exist.

Some progress has been made. The Louisiana Colonial Documents Digitization Project, funded by an $800,000 commitment from the Louisiana State Museum and a $275,000 NEH grant, has put 220,000 pages of French and Spanish colonial documents online at lacolonialdocs.org. The LSU Free People of Color project, backed by a $330,192 NEH grant, digitized 30,000 pages of family papers and business records from five institutions. PARES, the Portal de Archivos Españoles, offers free access to over 35 million digitized images from Spanish colonial archives, including the Archivo General de Indias in Seville.

These are real resources. They exist right now. Most Louisiana Creole researchers do not know about them.

AI can read what we could not

Here is where the story changes.

AI-powered handwriting recognition has reached a point where it can read 18th-century French cursive and Spanish colonial script with measurable accuracy. Transkribus, the leading tool, has over 300 models trained specifically on historical handwriting. On Spanish colonial manuscripts from the 16th and 17th centuries, it achieves a 5.25% character error rate for Redonda script and 8.92% for Italica Cursiva. Those are the exact document types sitting in Louisiana colonial archives.

Google’s Gemini model hit 0.56% character error rate on general handwriting, outperforming professional typists from the 1980s.

FamilySearch added 2.5 billion new searchable records in 2024. Their total is now over 20.5 billion. Ancestry’s AI processed 150 million records from the 1950 Census in nine days. The same task took nine months of manual work for the 1940 Census. MyHeritage launched Scribe AI in March 2026, adding transcription, translation, and document interpretation to their platform. BillionGraves deployed AI for gravestone transcription and hit 80% accuracy with half a million transcriptions in its first weeks.

The cost tells the real story. Human transcription of historical documents runs $2 to $5 per page. AI transcription runs $0.10 to $0.50. A 90% cost reduction. For a community organization with limited funding, that changes what is possible. A volunteer photographs a sacramental register at a parish church in Opelousas. AI reads the French cursive. Extracts the names, the dates, the relationships. A record that sat physically inaccessible for 200 years becomes searchable in hours.

The technology exists. The community does not know it yet.

AI also lies

I am not here to sell you on AI without telling you what is broken about it.

That is not a glitch. That is the system working exactly as designed, filling gaps with plausible fiction.

By late 2025, researchers documented over 729 cases of AI-generated hallucinations in court filings. Stanford found that legal AI models hallucinate in 1 out of 6 or more queries. Lexis+ AI hallucinated 17% of the time. GPT-4 at 43%. There is no reason to believe genealogical queries would be any more accurate.

For Black and Creole genealogists, this is not an abstract risk. We are already working with records full of gaps, erasures, and deliberate falsifications.

Standard OCR systems also show 23% higher error rates when processing documents about marginalized communities compared to mainstream records. The bias is baked into the training data. Retraining with diverse datasets reduces the disparity to under 5%, but that retraining has not happened for most Creole-relevant document types.

The Coalition for Responsible AI in Genealogy presented five principles at RootsTech 2026: accuracy, disclosure, privacy, education, and compliance. Their core message was simple. Make decisions based on your judgment. Not on what AI feeds you.

The census was never neutral

The census is the backbone of American genealogy. For Louisiana Creoles, it is also one of the most unreliable documents in the archive.

Census enumerators in Louisiana were political appointees. Untrained. Walking into Francophone Creole communities with no knowledge of the culture, the names, or the language. They spelled names phonetically. Arceneaux became Arsno, Arsnaux, Arsenaux, Arcino, and a dozen other variations depending on who was holding the pen.

Race was assigned, not reported. The enumerator looked at a person and decided. The 1930 Census instructions were explicit: a person of mixed white and Negro blood was to be returned as Negro, no matter how small the percentage. A Louisiana Creole family with French, African, Spanish, and Native ancestry got collapsed into a single box determined by one stranger’s perception on one day.

The mulatto category was dropped entirely in 1930. Self-identification was not introduced until 1960. Multiple-race selection was not allowed until 2000. For 70 years, the federal government had no category for what Louisiana Creoles had always been.

AI reading those records at scale is reading institutionalized bias at scale. It sees multiple people where there is one family. Or it confidently merges records that belong to different people because the names are close enough. Without someone who understands that Arceneaux and Arsno are the same person, that mulatto and white can be the same family ten years apart, the output is noise dressed up as data.

Who is missing from the room

The people who hold the keys to Black history do not look like the people whose history they hold.

The NIH funding gap for Black researchers has not closed since it was first documented in 2011. In 2025, fellowship grants to Black researchers dropped 40% while white researcher fellowships dropped 9%. Black scientists who study the topics that serve Black communities, health disparities and patient-focused interventions, receive lower scores from reviewers. The research that would close these gaps gets defunded first.

Gwendolyn Midlo Hall spent her career identifying 104,729 enslaved Africans from Louisiana records between 1719 and 1820. Her database is freely accessible at slavebiographies.org. Elizabeth Shown Mills wrote the foundational work on Cane River Creoles of Color. Ariela Gross at UCLA proved through legal analysis that law, not biology, constructed racial categories in Louisiana. These scholars did the work. The institutions have not caught up.

The opportunity is sitting right here

90% of Creole genealogical records are physical only. The records that exist are in colonial French and Spanish. AI can now read handwritten historical documents in both languages. The cost of transcription has dropped 90%. The NHPRC-Mellon program is actively funding African American history digitization with grants up to $60,000 per year, with priority given to HBCUs and minority-serving institutions. IMLS has awarded nearly $6 million for African American history digitization.

The money exists. The technology exists. The need is documented.

What is missing is someone on the ground in Louisiana who understands both the culture and the technology. Someone who knows that Arceneaux can be spelled eleven different ways. That the notarial act was a survival document. That the baptismal record was the only proof of freedom some families had.

This is where AI for Black Folks and Louisiana Creole Culture meet. Not in a conference room. In the archive. In the parish church. In the courthouse in Opelousas and the database in Seville.

The records are there. The tools to read them exist. A machine can now read 18th-century French cursive faster and cheaper than any human transcriptionist alive. That same machine can also invent ancestors who never existed and file them confidently into your family tree.

So here is the question I cannot answer for you.

Is AI the thing that finally opens 250 years of locked Creole records? Or is it the thing that fills the gaps with fiction and calls it genealogy? Is it the tool that gives a Creole family in Opelousas access to a baptismal register in Seville for the first time in 300 years? Or is it the tool that reads a census record written by a man who could not spell their name and treats his mistake as fact?

The honest answer is that it is both. Right now. At the same time.

What you do with that is yours.

Milton Arceneaux is a storyteller. The photography, the filmmaking, the brand work, the cultural advocacy, the executive tech career: all of it feeds the same thing. Telling the stories that need telling, with whatever tool fits. He founded Louisiana Creole Culture to fight the Cajunization of Louisiana history and defend what Creole of Color identity actually is. He runs Vues de Culture, its nonprofit arm, and created REFRAMING Cinema Film Festival to put minority voices on screen. He founded AI for Black Folks because he came from the tech side first and knows who these systems leave out. Encoded Noire, his creative agency, is the engine behind all of it.

I spent three hours one night Googling “grants for Black artists 2026.”

Here is what I got. A list from 2021. A Medium post linking to a dead page. Four different aggregator sites all pulling from the same broken database. A $29-a-month subscription wall asking me to pay to see “unlocked listings.” A foundation newsletter for a program that stopped funding the category two years ago. One actually relevant grant. Deadline had passed in November.

The aggregator trap

Most of what people call a “grant list” for Black creatives is a scraped database dressed up as curation. The sites make money from your email and your attention. They do not make money from you actually winning funding. That is not a partnership. That is a tollbooth.

I am an artist and creative director in South Louisiana. I run Encoded Noire, a Black-owned creative agency. I also run Louisiana Creole Culture, and I do creative work for a handful of small Black-owned businesses that do not have grant writers on payroll. None of us can afford to waste a Tuesday night clicking through listings that go nowhere.

So I stopped clicking and built my own.

What’s in the guide

Thirty-five grants. Every single one verified at the funder’s own website. Not scraped. Not copy-pasted from another guide. Not cached from 2022. If a listing could not be confirmed on the funder’s actual domain, it did not make the guide.

Six categories, because Black creative work is not one thing: federal humanities and state arts, documentary film and media, heritage, place, and preservation, small business and founder capital, AI, technology, and innovation, and French and Creole-specific funding.

Nineteen are open right now. Ten are on watch. Five closed recently and are already on the 2027 calendar so nobody has to go find them again next year. Every open grant inside the next 90 days is plotted on a visual deadline chart so you can see at a glance what is urgent and what can wait.

Every link goes straight to the source. Not a middleman. Not a referral page. The funder.

Why I am giving it away

Because the people who need this most are the ones most likely to have given up after hour two of that Google rabbit hole. Because research is expensive and my community has been paying that tax forever. Because I build brands for Black businesses, Black cultural organizations, and Black creatives for a living, and the same question comes up every single time — where is the money?

Here is some of it. Go get it.

This is also the first thing I am publishing through AI for Black Folks, and that matters. AIFBF exists to put tools in the hands of the people who got left out of the last tech wave. A free, funder-direct grant guide is exactly that kind of tool. Paywalls and $29 subscriptions are the opposite of it.

How I actually built this

One more note, because I want to be straight about the work.

Encoded Noire is a research-first creative agency. That sounds like a tagline but it is a working rule. Before we build a brand for a client, before we design a pitch deck, before we shoot a single frame, we do the homework. We scan the category. We find the competitors. We know where the money lives and who is already getting it. We do this for every engagement, not just the ones that impress.

This guide is what that rule looks like when I apply it to myself.

If you are a Black creative or a Black-owned business that needs the same rigor on a brand, a campaign, or a pitch deck of your own, that is what we do at Encoded Noire. But you do not have to hire us to use the guide. The guide is free. The email captures you to AIFBF, not to a sales pipeline.

Milton Arceneaux is the founder of Encoded Noire, a Black-owned creative agency in Louisiana, building brand identity, marketing strategy, photography, and AI-integrated creative work for clients across Acadiana and nationally.

Get the Free 2026 Grant Guide →