Frequently asked questions

Shannon Sheriff is an AI content strategist who plans, manages, and reviews content built for AI search. She has more than ten years of experience in content strategy and management, has been building with AI since January 2023, and writes about answer engine optimization (AEO), social media algorithms, and blockchain tokenization.

She grew a technology company's X audience from 3,800 to 90,000 followers in under six months, spent seven years as a content creator for Walmart, and is the author of NFTs: The Future is On Chain (2024). Full background.

AI content strategy is planning and managing content so it works for both people and AI systems. That means deciding what to publish and where, structuring it so answer engines like ChatGPT, Claude, Perplexity, and Gemini can understand and cite it, and using AI tools and agents to research and draft while people stay responsible for judgment, accuracy, and voice.

A good content review checks that the facts and numbers are right and sourced, that the piece answers the question its title promises, that it sounds like the brand and not like unedited AI output, and that it is structured so AI engines can pull a clear answer from it.

Four topics: how AI answer engines choose which sources to cite; how ranking algorithms on X, LinkedIn, Instagram, TikTok, YouTube, and Facebook are changing; building with AI agents as a non developer; and blockchain tokenization, including tokenized stocks, tokenized real estate, and AI agent wallets.

NFTs: The Future is On Chain, published in 2024. It explains NFTs and on chain ownership for readers coming to blockchain from outside the industry.

On X at @shannonsheriff, where she posts AEO, algorithm, AI, and tokenization updates. Longer pieces are published on shannonsheriff.com/blog, which has an RSS feed. She is also on LinkedIn.

Yes. She has been building with AI since January 2023. She is not a developer. She learns by building, using plain English instructions, and writes about what worked and what broke in public.

A chatbot answers questions. An agent takes actions. An AI agent can be given access to tools such as files, email, a browser, or a calendar, and can carry out multi step tasks on a schedule. The difference is like asking someone for advice versus asking someone to actually do the thing.

Yes. Tools such as OpenClaw let you run an agent on your own computer and talk to it through a messaging app like Telegram. You describe the task in plain language, give the agent only the access it needs, such as one inbox or one folder, and ask it to confirm before it deletes or sends anything. A simple first project is a repeating chore, like clearing a certain kind of spam or sorting downloads, that the agent can then run on a schedule.

It works as a long game. Over one week in late October 2026, 21,928 posts on X mentioned building in public, and most produce little. But public work has turned directly into job offers and revenue: Peter Steinberger went from a weekend open source project to joining OpenAI in about three months, and Aurora Scharff joined Vercel after years of public React and Next.js teaching. The common thread is consistency over months, not one viral post. Read the full analysis.

An AI search audit looks at a site the way crawlers and answer engines do. It checks that AI crawlers are allowed in robots.txt, that structured data correctly names the person or business behind the site, that each key page answers a clear question near the top, that headings and page structure are clean, that pages load quickly without relying on scripts to show text, and that the name, role, and facts match across the site, social profiles, and directories.

Answer engine optimization is the practice of making a person, brand, or page easy for AI systems to understand, trust, and cite when they write an answer. The goal is to be named or linked inside the response from ChatGPT, Claude, Perplexity, Gemini, Copilot, Grok, Meta AI, or Google AI Overviews, not just to appear in a list of links.

SEO (search engine optimization) works to rank pages in a list of search results, mainly through keywords, backlinks, and technical health. AEO works to get cited inside an AI generated answer, mainly through clear entity signals, direct answers to real questions, structured data, and facts that match across the web.

The two overlap. AI engines that browse the web still rely on search indexes, so a site that cannot be crawled or ranked is hard to cite. Short version: SEO gets you ranked, AEO gets you cited.

Very little in practice. Both describe getting surfaced by AI systems. GEO (generative engine optimization) is the term used in academic research and focuses on how content shows up inside AI generated text. AEO (answer engine optimization) is more common in marketing and focuses on being the source an AI picks when it answers a question. The work is the same: clear facts, clear authorship, structured pages, and a consistent presence across the web.

When an AI assistant searches the web, it retrieves a set of pages and picks the passages that answer the question most directly. Pages tend to get cited when they state the answer plainly near the top, come from an author or organization the system can identify, agree with what other reputable sources say, are recent, and can be crawled without logins or heavy scripts.

Without a live search, a model answers from what it learned in training, which favors names and facts that appear consistently across many sources over time.

An entity is a specific, identifiable thing: a person, company, place, or product. Entity authority is how confidently search engines and AI systems can connect that entity to a topic. It grows when the same name, role, and expertise show up consistently across your website, social profiles, publications, and mentions by others, and when structured data links those profiles together.

It helps machines read your page correctly. Schema.org markup in JSON-LD states facts such as who wrote a page, who a person is, and which profiles belong to them. It does not guarantee a citation, but it removes guesswork, and it is one of the few signals you control completely. Person, ProfilePage, Article, FAQPage, and Organization are the most useful types for personal and brand sites.

llms.txt is a proposed standard, introduced in 2024, for a plain markdown file at the root of a website that summarizes the site and links to its most important pages for large language models. It is cheap to add and easy to keep current, but support from the major AI companies is uneven, so treat it as a helpful extra rather than a ranking factor. This site has one at /llms.txt.

Only if you do not want to appear in AI answers. Most AI companies run separate crawlers for model training and for live search:

OpenAI uses GPTBot for training and OAI-SearchBot for ChatGPT search. Anthropic uses ClaudeBot and Claude-SearchBot. Perplexity uses PerplexityBot. Google uses the Google-Extended token to control use of content for Gemini. Meta uses Meta-ExternalAgent to train its AI models and index content, Meta-ExternalFetcher to fetch a page when a user asks Meta AI about it, and FacebookExternalHit to build link previews on Facebook, Instagram, and WhatsApp.

Blocking a search or fetch crawler removes you from those answers, and Meta notes that its fetcher and preview crawler can bypass robots.txt in some cases. Check your robots.txt before assuming AI cannot see you.

Create a custom channel group in Google Analytics 4 with a channel called AI Search, and match the session source against the AI domains, for example: chatgpt.com, perplexity.ai, claude.ai, gemini.google.com, copilot.microsoft.com, meta.ai. Place it above the Referral channel so those visits are not lumped in with ordinary referrals.

Some AI traffic still arrives as Direct because apps do not always pass a referrer, so treat the number as a floor, not a total.

Shares, above all copy link shares. In the For You ranking weights X published in August 2026, a predicted copy link share carries a weight of 20.0, the highest positive action in the file. A reply, a quote, and a share by DM each carry 5.0. A like carries 0.5.

The weights multiply the model's predicted probability of each action, so they are not an exact exchange rate between likes and shares. The takeaway still holds: write posts people want to send to someone, and ask questions people want to answer.

Constantly. Every major platform adjusts ranking continuously and announces only some of it. The useful habit is tracking what each platform says it is optimizing for, such as shares, replies, watch time, or saves, and checking your own analytics after each announced change. Shannon Sheriff posts algorithm updates for X, LinkedIn, Instagram, TikTok, YouTube, and Facebook on X.

Asset tokenization is representing ownership of, or exposure to, an asset as a token on a blockchain. The asset can be a stock, a bond, a fund share, real estate, or a commodity. Tokens can trade outside traditional market hours, settle quickly, be divided into small fractions, and be used inside on chain applications, but the rights a holder actually gets depend on how the token is structured.

Real world assets, or RWAs, are traditional financial and physical assets that have been brought on chain as tokens. Common examples are tokenized U.S. Treasury funds, private credit, stocks, real estate, and commodities like gold. RWA is one of the fastest growing categories in crypto because it connects blockchain markets to assets with established cash flows.

Real estate tokenization turns ownership of a property, or a share of the income from it, into tokens on a blockchain. In most U.S. deals the property is held by a company such as an LLC, and the tokens represent shares of that company rather than the deed itself, which is still recorded with the county.

The appeal is fractional ownership, lower minimums, and the potential to sell a stake without selling the building. The tradeoffs are real: these tokens are usually securities, so offerings follow securities rules and are often limited to certain investors, and a token is only as liquid as the market where it is allowed to trade.

The same two ways as traditional property investors: a share of rental income, paid out as distributions, and a share of any gain when the property is sold or refinanced. Some platforms pay distributions in stablecoins. Returns depend on the property, the fees, and the legal structure behind the token, so reading the offering documents matters more than the blockchain it runs on.

No. The term covers at least four structures. Synthetic tokens and perpetual futures track a stock's price with no share behind them. Issuer backed tokens are supported one to one by real shares, but the holder's claim is on the issuer, not the share. Direct claim tokens give the holder a claim on a share held at a regulated custodian. Tokenized securities are the share itself, with the same economic and voting rights.

The questions that matter are what backs the token, who holds the share, whether you can redeem it, which rights pass through, and where and when it trades.

On September 17, 2026 the SEC issued a five year conditional exemption that lets approved venues trade tokenized versions of U.S. listed stocks on chain. It covers tokens that carry the same economic and voting rights as the stock, not tokens that only track the price, and it comes with conditions such as volume caps and no leverage.

A tokenized stock is meant to represent a share or a claim tied to a real share. A stock perpetual is a derivative contract that tracks the share price with no expiry and no share behind it. Perpetuals can offer leverage and trade around the clock, but holders get no dividends, votes, or ownership, and they carry counterparty and funding rate risk.

An AI agent wallet is a crypto wallet that an AI agent can use to act on chain, such as making trades or payments, within limits its owner sets. It is where AI agents and tokenization meet: software that can hold and move tokenized assets on someone's behalf. Shannon Sheriff is a member of the MetaMask Agent Wallet Traders beta group, which tests this kind of setup.