# AI influencer traffic: Instagram reach under new rules AI influencer reach on Instagram: how the AI-generated profile label works, how to spot a throttle, how to post safely, and how to use Reddit and TikTok. - Canonical: https://aiofmcity.com/guides/traffic-and-platform-policy - Topic: Traffic & policy (https://aiofmcity.com/topics/traffic-and-policy) - Author: Gus - Published: 2026-09-08 - Updated: 2026-10-21 --- ## Key points - Since 31 August 2026, an AI-generated profile on Instagram without the label reaches fewer non-followers. - One bad post does not mean you are throttled. Look at how many non-followers you reach over several weeks. - Platforms do not look for AI content. They look for accounts that behave like bots. - On Reddit, use two accounts: one to be helpful as an operator, and a separate one for persona content. - Write down your own numbers every week. Platform analytics only show a short window, and you need the trend. --- Most AI persona projects fail at getting traffic. The usual reason is simple. You built your whole workflow around how one platform behaved, and then the platform changed. The rules change about once a quarter. This guide shows you how to set up your traffic so that the next rule change does not break it. ## Instagram's new AI-generated profile label The biggest change of 2026 for AI accounts was not an algorithm change. It was a new labelling rule. > On 31 August 2026 Instagram replaced the "AI creator" label with an "AI-generated profile" label. Accounts whose depicted identity is AI-generated and that do not set the label receive reduced distribution to non-followers on recommendation surfaces. Accounts that merely use AI as a tool for editing or captions are not covered. > > — Source: Instagram Creators — AI-generated profile label, read 2026-09-08 — https://creators.instagram.com/blog/ai-generated-profile-label This means three things for you. The [action plan for the label](https://aiofmcity.com/blog/instagram-ai-generated-profile-label) goes through each of them. 1. The label is not a punishment. Setting it is how you follow the rule. You lose reach when you do not set it. 2. The reach you lose is reach to people who do not follow you yet. A growing account depends on exactly those people. 3. The rule looks at who the account shows. If the person in your posts is AI-generated, the rule applies to you. No caption wording changes that. ### How to check if your reach is already limited Do not wait for Instagram to tell you. Open Account Status and check. That is where Instagram states the reason, so you do not have to guess it from a chart. > Creators can check in Account Status whether their account is no longer eligible for recommendations, and can file an objection against a restriction they believe was made by mistake. If the objection succeeds, the profile and content are displayed in recommendations again. > > — Source: Instagram Creators — AI-generated profile label, read 2026-09-09 — https://creators.instagram.com/blog/ai-generated-profile-label Insights has no number that shows this restriction. Your reach to non-followers can drop for many other reasons too, and Insights does not tell them apart. So check Account Status every week, together with the weekly numbers described below. If you want to object, [Instagram AI-label appeal](https://aiofmcity.com/blog/instagram-ai-influencer-banned-appeal) explains what a good objection contains. You also need to know which accounts the rule covers. > The rule applies only to profiles whose depicted identity is AI-generated. Using AI as a tool — retouching, upscaling, background replacement, caption writing, scheduling — does not put an account in scope. > > — Source: Instagram Creators — AI-generated profile label, read 2026-09-08 — https://creators.instagram.com/blog/ai-generated-profile-label The test is one question: is the person shown on the account AI-generated? A fully generated persona is covered, even if almost no AI goes into its captions. A real model whose photos are retouched with AI is not covered, even if the whole workflow behind the account is automated. Do not set the label "just in case" on an account it does not describe. The label would then tell Instagram's recommendation systems something false about the profile. The label article has the full table of edge cases. It includes mixed accounts, which have a consent problem before they have a labelling problem. ## What throttled reach looks like You do not get a message when a platform shows your posts to fewer people. You have to work it out from your numbers. Most people get it wrong, because they compare one post with the post before it. Measure the **non-follower share of reach** over several weeks. That is the part of your reach that comes from people who do not follow you. If you are throttled, that share falls and stays down, while your followers engage about as much as before. One weak post proves nothing. A month of weak posts with normal follower engagement is a pattern. Write the numbers down yourself every week, in a place you control. Platform analytics only show you a limited time window. You need the long-term trend, and the platform does not keep it for you. _Diagram: Judge reach by the non-follower share over several weeks. If it falls while your followers respond as before, it is a throttle; if everything dips together, it was a soft week._ ### What did not work: Diagnosing a throttle from a single bad week Reach fell sharply on three posts in a row. I decided the account had been throttled. I changed the posting times, rewrote the captions and cut the schedule. Reach came back, and I took that as proof that I had been right. I was wrong. Later I went through the numbers I had kept for that month. Follower engagement had dropped in exactly the same week. It was a weak week for everything, and everything recovered at the same time. I had read a random dip as a throttle. Then I read the end of the dip as proof that my changes had fixed it. I lost three weeks changing things that were working. ## How to post without looking like a bot Platforms are not looking for AI content. They are looking for accounts that behave like bots. You can post AI content without behaving like one. - **Post at times a person would post.** Do not post at the same minute every day. Do not do the same action many times in a row. You can use a scheduling tool. The warning sign is a schedule that is as regular as a machine. - **Write real replies.** Sending the same text many times is the fastest way to look like a bot, because bots do exactly that. If you cannot keep up with the replies, reply to fewer people. Do not switch to templates. - **Give the persona one voice and keep it.** A single caption can look fine on its own. But if all your captions have the same rhythm, the account as a whole reads as generated. [Where the hours go](https://aiofmcity.com/blog/where-the-hours-go-ai-persona) explains why captions are the wrong thing to automate first. - **Label the account as AI everywhere you post.** That is how you follow the rules. It also means that a reviewer who looks at the account finds nothing hidden. ## TikTok: you label each video Everything above is based on Instagram's rules. The most expensive mistake is to assume that the next platform has the same rules. TikTok has no label for the whole account. It has rules for each video, and it reads the data stored in the file. > By May 2024, TikTok had implemented C2PA Content Credentials to automatically recognise and label AI-generated content — including content made on other platforms and uploaded to TikTok — and began attaching Content Credentials to content generated with TikTok's own AI effects, so the metadata survives a download. > > — Source: TikTok Newsroom — Partnering with our industry to advance AI transparency and literacy, read 2024-05-09 — https://newsroom.tiktok.com/en-us/partnering-with-our-industry-to-advance-ai-transparency-and-literacy If you post to both platforms, three things change: 1. You label each video. On Instagram, one setting covers the whole profile. On TikTok, every realistic video needs its own label, and you set it before you post. If you miss a label, the problem is with that one video and not with the whole account. 2. TikTok can label the video for you. If the file still carries its provenance data (the record of how it was made), TikTok may add the AI label itself. You then find out when you see the label on your video. 3. How you export the file matters. It decides whether the provenance data is still there for TikTok to read. [Don't strip your watermarks](https://aiofmcity.com/blog/dont-strip-your-watermarks-provenance) explains what to check. [Does TikTok allow AI influencers](https://aiofmcity.com/blog/does-tiktok-allow-ai-influencers) shows how the rule developed. The [TikTok page](https://aiofmcity.com/platforms/tiktok) has the current version and every change with its date. ## X and YouTube: AI labels, but no rule for AI personas X and YouTube are different again. Both show a "Made with AI" label. Neither has a rule for an account whose whole identity is generated. YouTube's AI rule talks about real people, places and events. It does not say what applies to a fully invented character. In March 2026, X added a label that creators set themselves. X's binding rules still do not say that you have to use it. The only X rule that touches personas is about fake personas used to deceive people. A platform with no rule has not given you permission. These are also the platforms where a new rule is most likely to come next. So do what this guide already recommends. On every platform, label the media and say in the bio what the account is, even if the platform does not ask for it. When the next rule arrives, you only have to switch on a setting. You do not have to go back and fix all your old posts. The [platform rules](https://aiofmcity.com/platforms) pages have the current wording for each platform. The [label matrix](https://aiofmcity.com/ai-label-matrix) puts them side by side. YouTube has its own label and its own monetisation line: [AI influencer YouTube channel](https://aiofmcity.com/blog/ai-influencer-youtube-channel). _Where the AI label sits on each platform (checked 2026-09-08; full wording and sources: https://aiofmcity.com/ai-label-matrix)_ | Platform | Label on the account | Label on each post | Labels from file metadata | Note | | --- | --- | --- | --- | --- | | Instagram | required | for some content | yes | Post label only for realistic video or audio | | TikTok | none | required | yes | Label each realistic video before you post it | | YouTube | none | for some content | yes | The rule is about real people and events | | X | none | optional | partly | You can set it, but no rule says you must | | Reddit | none | none | not documented | Each subreddit decides | | Fanvue | required | required | not documented | Account tag plus a label on the content | ## Getting fans from your social account to Fanvue Reach alone does not earn you anything. You earn when people move from your social account to your paid account. One rule for that step is worth more than any posting tactic. > Fanvue advises putting the link in your bio on every platform you post to, and > never posting a bare Fanvue link directly in a caption or comment — routing > through a bridge such as a link provider or a custom tracking URL instead, > because a bare link risks getting the social account suspended. It also > recommends a separate tracking link per platform, and comparing which one > converts rather than which gets the most clicks. > > — Source: Fanvue Help — Getting Fans to Your Fanvue: Socials & Link-in-Bio, read 2026-09-09 — https://help.fanvue.com/en/articles/16238089-getting-fans-to-your-fanvue-socials-link-in-bio Take the first part seriously. The most common way to lose a persona account is to paste the subscription link into a caption. Missing labels and reported posts are less common causes. Several platforms treat a pasted link as off-platform solicitation, which means advertising a paid service that sits on another site. The bridge page does more than count clicks. It protects your social account from a ban. On Instagram, the paid Meta One Advanced plan adds clickable links in up to four posts and four reels a month. [Can AI influencers get verified on Instagram?](https://aiofmcity.com/blog/ai-influencer-verified-instagram) shows what the plan gives a persona account and what it costs. The second part is about what you measure, and many people measure the wrong thing. Clicks show you which platform sends the most visitors. Conversions show you which platform sends people who subscribe. Often these are two different platforms. Say one platform brings a tenth of your clicks and half of your subscribers. Then put your next month of work into that platform. **Where you have to say it is AI at each step.** Even people who set up the funnel correctly get this wrong. Each place has its own rule, and following one rule does not cover the others: - **On the social account**, use the platform's own label. On Instagram it covers the whole account. On every other platform the label has a different name, sits in a different place and applies in different cases. The [label matrix](https://aiofmcity.com/ai-label-matrix) lists them platform by platform. [TikTok specifically](https://aiofmcity.com/blog/does-tiktok-allow-ai-influencers) has rules for the video and not for the account, so a mistake there affects one video. - **On the bridge page**, the law does not ask for anything. It is still the page most likely to contradict your label. If the profile is labelled as AI and the landing page presents the persona as a real woman, you still contradict yourself, only in a different place. - **On the subscription profile**, follow Fanvue's own rule: say it is AI in the bio, in the caption or with a watermark. - **On the media itself**, Article 50 of the EU AI Act applies, wherever the file ends up. This rule stays with the file. None of the steps above covers it. The easy way is to write one sentence that says the persona is AI, and to use it in all four places. Do not use a discreet version on one platform and a clear version on another. A moderator can spot that contradiction in thirty seconds. ## Reddit is two different jobs Reddit is where the operators in this field talk to each other. It is also where persona content gets removed fastest. If you mix the two, you get banned from the communities you need. - **Being useful in operator communities.** Do this under your own operator identity, never as the persona. This is where you get answers to your questions, and it is where the content on this site comes from. Do not promote anything there. It stops working the moment you do. - **Posting persona content.** Use a separate account for this. Read the rules of each subreddit one by one. Expect this account to live much shorter than the first one, and plan for that. Never let it touch the first account. [Reddit for an AI persona](https://aiofmcity.com/blog/reddit-ai-persona-promotion) explains what gets an account banned there. It is not the self-promotion ratio you have probably heard about. Reddit's own rule is short. All the other rules come from each community. > Reddit permits AI-generated content, subject to each community's specific rules. The prohibition that applies platform-wide is AI content that "presents itself as human-generated." There is no platform-wide AI label, toggle or flag — disclosure, where a community wants it, happens in the post itself: a tag in the title, a flair, or a line in the body. > > — Source: Reddit Help — Manipulated Content and Misleading Behavior, read 2026-05-19 — https://support.reddithelp.com/hc/en-us/articles/41180423371156-Manipulated-Content-and-Misleading-Behavior So Reddit itself has one rule, and the moderators decide everything else. One subreddit wants a flair on every post. Another bans generated images. Another limits links. Each subreddit writes its rules in its sidebar, and you only know them if you read them before you post there. A persona account that says it is AI follows Reddit's own rule everywhere. You still have to check the rules of each subreddit separately. The two-account split also protects you, because of what Reddit says it now looks for. > Reddit reports blocking 23 million spam views a day and catching roughly 25,000 new spam posts and comments daily, a 20% reduction in user exposure to spam in the most recent quarter it reported. It says it uses LLMs specifically to catch "the highly subtle, coordinated patterns of fake behavior" that earlier systems missed — checking for suspicious signals from the moment an account is created. > > — Source: TechCrunch — Reddit is using LLMs to solve a problem LLMs largely created, read 2026-07-06 — https://techcrunch.com/2026/07/06/reddit-is-using-llms-to-solve-a-problem-llms-largely-created/ This is exactly the kind of pattern Reddit means: one account posts persona content, and another account upvotes it, comments on it or links to it. You avoid the pattern if you really run the two accounts apart. Use them at different times. Never let one vote for the other. Never post the same link from both within minutes. If you run both accounts from one session in one sitting, you produce the pattern, whatever the accounts are called. ## How to prepare for the next rule change The Instagram label was not the first rule change that broke working setups in this field, and it will not be the last. Three habits turn the next change into a small adjustment: _Diagram: Three small habits. If you skip one, the next policy change can turn into a crisis._ - Own your audience — reach them off the platform - Read rules every quarter — write down the date - Write the plan — before anything breaks Make sure you can reach your audience somewhere you control. If your audience exists only on one platform, you lose it when the account is banned or the algorithm changes. Read each platform's policy pages once a quarter, and write down the date you read them. If you miss a change, it can cost you reach. Write your recovery plan while nothing is wrong. [Risk, bans and recovery](https://aiofmcity.com/guides/risk-bans-and-recovery) is the guide for that. The plan should fit on one page. ### What did not work: Building the posting workflow around one platform's current behaviour I tuned the workflow to what worked on one platform in one quarter: post format, timing and caption structure. It worked well. Then the platform changed its labelling rules. I had no second platform to fall back on, and I had no record of what the rules had been before. So I could not tell how much of the drop came from the new policy and how much came from me. Rebuilding took longer than building had, because I did it while reach was already down. What I learned applies to every platform. A workflow that depends on one platform fails when that platform changes. I had built it that way on purpose, because it was faster. The articles in [Traffic & policy](https://aiofmcity.com/topics/traffic-and-policy) cover each rule change when it happens. The rules change every quarter. The habits on this page stay the same. ## Frequently asked ### Does labelling the account as AI-generated reduce reach? On Instagram, the missing label is what costs you reach. Since 31 August 2026, an account whose depicted identity is AI-generated and that has no label is shown less in recommendations. Setting the label is how you follow the rule. Instagram has not said whether labelled accounts are shown differently from human accounts. No one outside Instagram can answer that. ### How do I know if I have been throttled rather than just having a bad week? Look at how many non-followers you reach over several weeks. Do not compare one post with the next. If you are throttled, the non-follower share drops and stays down, while your followers engage as before. One weak post means nothing. ### Is Reddit worth the effort for an AI persona? Yes, if you use it to be helpful in the communities where operators talk. Posting persona content is different: Reddit is where those accounts get removed fastest. Use two separate accounts, because each use has its own rules. ### Should posting be automated? You can schedule posts. Do not let the pattern look automated. Platforms look for posts at machine-regular times and for the same reply text sent many times. Better software does not fix that. Post at the times a person would post.